03-Titel-1 + Bold Master of Occus

Home , British nationalism, Caledonian Union, Flemish Movement, Inuit Party, Motherwell North (ward)

03-titel-1 + bold Master of Occus

03-titel-4 As aliquia natum quo ea que quiae cum rorae.

Regional Resources and Democratic Secessionism

Kai Gehring & Stephan A. Schneider

CIS Working Paper No. 90

Center for Comperative and International Studies (CIS) Regional Resources and Democratic Secessionism

Kai Gehring ∗

Stephan A. Schneider ∗∗

31st October 2016

Abstract: We show that economic concerns revolving around region-specific resources are crucial to understanding the success and failure of democratic secessionist movements. Based on a simple voting model, qualitative evidence from numerous countries indicates a clear relationship between the value of regional resources and the electoral success of the respective regionalist parties. We then make use of a natural experiment in the United Kingdom, where the discoveries of oil off the Scottish coast allow us to implement a difference- in-differences design with Scotland as the treatment group and Wales as a suitable counterfactual. Using election results at the constituency level over the period from 1945 to 2001, we document a significant positive effect on the vote share of the Scottish National Party of about 2 percentage points for an additional giant oil discovery. The causal interpretation of this effect is further supported by exploiting the exogenous change in oil prices in a triple-differences design.

Keywords: Secession, Separatism, Size of Nations, Resources, Nationalism, Economic Voting JEL Codes: H77, N9

Acknowledgments: We thank Toke Aidt, Alberto Alesina, Bruno Caprettini, Axel Dreher, Florian Foos, David Hemous, Jan-Egbert Sturm, Hans-Joachim Voth, and the members of Clare Hall at the University of Cambridge for comments, as well as participants at the PEUK Conference in Oxford 2016, the EPCS 2016 in Freiburg, and the Development Conference “Goettinger Schule” in Göttingen, and seminars at the political science and the economics department of the University of Zürich. We also thank Christiane Baumeister, Myron K. Horn, and Pippa Norris for sharing their data as well as James D. Hamilton and Valerie A. Ramey for valuable suggestions. Valerie Baldinger, Janine Albiez, and Michael Andrews provided excellent research assistance and support. Stephan Schneider conducted part of this research at the University of Cambridge with ﬁnancial support from the Cusanuswerk.

∗ Kai Gehring, University of Zürich: [email protected]. Webpage: http://www.kai-gehring.net. ∗∗ Stephan A. Schneider, Heidelberg University: [email protected]. 1 INTRODUCTION 1

“The foolish ones said to the wise, ‘Give us some of your oil; our lamps are going out.’ ‘No,’ they replied, ‘there may not be enough for both us and you. Instead, go to those who sell oil and buy some for yourselves.’ ”

Matthew 25, 8-9 – The Parable of the Wise and Foolish Virgins

1 Introduction

Many regions around the world have experienced a surge in the vote share and success of secessionist parties, while others with comparable levels of cultural heterogeneity have not. This cannot be explained by the existing literature on the optimal size of nations which has mostly focused on scale advantages like improved trade opportunities as the main benefits from larger states, and preference heterogeneity as the most severe factor benefiting secession and smaller entities. Those factors are obviously important, but we argue that regional resources and their changing value are crucial to explain the success of secessionist parties. This connects to an existing literature which shows that in developing countries increases in natural resource wealth in individual regions affect state stability and democracy (see, e.g., Arezki & Brückner, 2012, 2011). The emerging distributional conflicts often culminate in armed conflicts and severe civil wars (e.g., Collier, 2010; Rosser, 2006; Ross, 2004a), which is fortunately unlikely in countries with a stable democratic system. Instead, we argue that secessionist tendencies find their expression in a different way: the redistribution and usage of the gains from regional resources can be instrumentalized by secessionist parties to boost their vote shares.

Our simple probabilistic voting model builds on the assumptions and relevant factors highlighted by authors like Alesina & Spolaore(1997) and augments them by integrating differences in resource endowment and their redistribution across regions. This also relates our study to the literature on fiscal federalism and within-country transfers (e.g., Dixit & Londregan, 1998; Persson & Tabellini, 1996). We use extensive anecdotal evidence to demonstrate that changes in resource value are decisive in influencing the rise and fall of secessionist movements around the world. Ranging from Flanders in Belgium, to Québec in Canada, and Greenland in Denmark, the cases we describe document the widespread relevance of regional resources. Moreover, they demonstrate that regional resources should not narrowly be defined as containing only natural resources, but rather that regional differences in human capital, institutional quality or geography are also relevant. Based on Alesina & Spolaore(1997) and Persson & Tabellini(2002), we derive analytically that increases in the value of regional resources is related to a higher approval rate for secession, which should be visible, ceteris paribus, in a higher vote share for secessionist parties.

Based on the model, we identify the challenges and necessary prerequisites for identifying a causal effect of regional resources. While the qualitative examples underline the general importance and external valid- 1 INTRODUCTION 2 ity of the hypothesized mechanism, they only provide anecdotal and correlational evidence how regional resources and changes in their value help to fuel secessionist movements. Ideally, we want to compare two regions that exhibit comparable trends with regard to those factors that we identified as influencing support for secession; for instance, preference heterogeneity from the majority in the country. Moreover, both regions must feature a comparable secessionist party and the vote share of these parties is required to follow a common trend in absence of the treatment. We identify such a case in the United Kingdom and make use of the fact that the bulk of the UK’s North Sea oil reserves was exclusively found off the Scottish coast to compare the performance of the two major secessionist parties in Scotland and Wales in a difference-in-differences (DiD) framework. This is feasible as extensive evidence supports the notion that the Scottish National Party (SNP) and the Welsh Plaid Cymru as well as both regions in general are sufficiently comparable for our purpose (see, e.g., Levy, 1995). Moreover, the first Scottish oil discoveries in 1970 were an unexpected and considerably large exogenous shock (MacKay & Mackay, 1975), which makes them salient to voters and allows a clear distinction between pre- and post-treatment period.

Our analysis, based on a panel dataset containing 1883 observations from UK general elections and by- elections on the constituency-level which we assembled for the 1945 to 2001 period, exploits this natural experiment. While oil did not play an important role in the political calculus of secessionist parties in Scotland and Wales initially, the situation fundamentally changed for the Scots by the initial oil discoveries in ‘their’ maritime area. Hereafter, politicians from the SNP have tried to instrumentalize the large potential oil revenues as an argument for Scottish independence, and support for the party has increased sharply (McGuinness et al., 2012).1 The slogan “It’s Scotland’s oil”, invented in 1972 and often quoted even today (Harvie, 1995), reﬂects concisely how politically relevant the oil discoveries were and still are for the SNP (Collier & Hoeﬄer, 2006; MacKay & Mackay, 1975).

Our results show that voters are indeed receptive to these arguments: in elections after oil discoveries, the SNP’s vote share significantly increased by about two percentage points for each additional giant oil discovery. The result is robust to various different specifications, is further supported by exploiting the exogenous changes in world oil prices and a placebo test, and varies with constituency characteristics to some degree. This augments the qualitative evidence that we collected for numerous other countries, and highlights that regional resources are a crucial factor to better understand the success and failure of secessionist parties.

1 Support for independence provisionally culminated in the referendum held on 18th September 2014, in which the Scottish people faced the choice of separating from the rest of the United Kingdom or staying uniﬁed. Even though a close majority of the Scottish people voted against independence, the fact that an independence referendum was held in a stable West-European country with a wealthy society is remarkable in itself. The oﬃcial electoral outcome of the referendum is as follows; yes: 1,617,989 (44.5 percent); no: 2,001,926 (55.5 percent); turnout: 84.6 percent. The question asked on the ballot paper was: “Should Scotland be an independent country?” (see http://www.scotreferendum.com/information/, last accessed on 31st October 2016). But Scotland is no isolated case: in Québec, for instance, an independence referendum closely failed the majority in 1995 (Holitscher & Suter, 1999; Lynch, 2003). 2 LINKING RESOURCES AND SECESSIONISM 3

The paper is structured as follows: section2 relates our contribution to the existing literature on secessionism in developing countries and the size of nations as well as presents our theoretical model. We demonstrate the importance of regional resources and the external validity of our ﬁndings with detailed anecdotal evidence from various secessionist parties around the world (3.1). Subsequently, we provide a brief overview of the historical and political background of the Scottish and Welsh independence movements (3.2), and develop the hypotheses tested in part5. Sections 4.1 and 4.2 provide a comprehensive data description and details about the identiﬁcation strategy. Section5 then presents regression results and discusses the robustness of the estimates. Section6 concludes.

2 Linking Resources and Secessionism

2.1 The Political Economy of Resources and Conﬂict

There is a large strand of literature which examines the positive and negative economic consequences of resource endowments and discoveries for economic development in the context of developing countries (for an overview see, e.g., Caselli et al., 2015; Ploeg, 2011; Rosser, 2006). Numerous studies focus specifically on political consequences and suggest that an abundance of natural resources leads to more corruption, lower levels of democracy, and civil war (see, e.g., Rosser, 2006). Secessionist civil wars represent a violent form of secessionism which often arises due to distributional conflicts about resources (see, e.g., Ross, 2004b; Collier & Hoeffler, 2004).2 Lei & Michaels(2014), for instance, show that the discovery of oil enhances the probability of internal armed conflicts by 5 to 8 percent within 4 to 8 years after the respective discovery.

In addition to grievances and oppression in the respective regions, various authors highlight the relevance of an economic calculus to explain the onset of a civil war. Lei & Michaels(2014) and Morelli & Rohner (2015) argue that the winner of a civil war gains control over the resource repositories in the contested area. If these expected revenues are sufficiently high and the expected costs sufficiently low, a secessionist conflict is likely to arise. Collier & Hoeffler(2004) support this so-called greed hypothesis: their results indicate that beyond ethnic differences, potential separatist movements in developing countries also need prospective economic gains from separation. A rebellion is more likely to be initiated when it is both financially feasible and potentially rewarding (see also Collier & Hoeffler, 2006; Ross, 2004b).

Still, it appears as if it is neither only greed nor solely grievance which fuels secessionist conﬂicts. As we will argue in the context of democratic countries below, ethnic, cultural or linguistic diﬀerences are

2 Anecdotal evidence for this relationship is abundant: examples of civil wars related to resource wealth can be found in Angola, Colombia, Iraq, Sudan, Indonesia, Nigeria, or the Congo (Casertano, 2013; Lei & Michaels, 2014). 2 LINKING RESOURCES AND SECESSIONISM 4 important factors that can affect the emergence of secessionist movements. However, the degree to which they effectively carry over into real actions is often depending on the economic circumstances revolving around resource distribution. This becomes apparent in the study by Morelli & Rohner(2015). When interacting resource and ethnicity concentration, they demonstrate that a civil war is more likely to occur when both factors are densely concentrated (see also Casertano, 2013; Collier & Hoeffler, 2006). More specifically, resources need to be geographically concentrated and the different groups within the country need to be distinct enough to enable group leaders to instrumentalize the unequal distribution for their purpose.

Economic arguments are often plausible mechanisms to explain separatist wars because they increase the perceived value of independence in the resource rich region (Ross, 2004a). 3 Morelli & Rohner(2015) also show that a shift from full oil equality to full oil inequality, ceteris paribus, quadruples the baseline risk of civil war. Thus, we can learn from the developing country context that the distribution of region-specific resources in the respective state is a crucial factor causing secessionist wars. We argue that the framework applied to the context of the onset of civil war can be adapted to reflect the case of non-violent secessionism as well. This idea is supported by Collier & Hoeffler(2006), who already hint at the possibility to transfer the knowledge about secessionist wars to non-violent secessionist movements. They also qualitat- ively discuss the potential relationship between the Scottish independence movement and geographically concentrated oil discoveries off the Scottish coast as a prime example of democratic secessionism Collier & Hoeffler(2006).

2.2 Economic Voting and Secessionism

2.2.1 Beneﬁts and Costs of a Union

In his book on the relationship between the economy and electoral outcomes, Tufte(1978, p. 65) appeals to the reader: “When you think economics, think elections; When you think elections, think economics.” On average, economic changes explain about one third of the vote (Lewis-Beck & Stegmaier, 2000). 4 Analyses

3 This claim fits empirical evidence and case studies. Explaining possible chains of effects, Ross(2004a) provides case study evidence for three separatist wars where a resource rich region wanted to secede. In a sample of 127 separatist wars between 1945 and 1999, Popiden(2011) demonstrates a positive relationship between dependence on the export of non-renewable energy resources and the onset of separatist wars. In addition, she shows that dependence on oil exports is a greater risk factor for the onset of a separatist war than a non-separatist war. Results presented by Collier & Hoeffler(2006) also support the notion that natural resources rather fuel secessionist conflicts than ideological wars. 4 The three indicators unemployment, growth, and inflation are generally deemed as the most important since they are among the few publicly recognized economic indicators (see, e.g., Lewis-Beck & Stegmaier, 2000, p. 114; Powell & Whitten, 1993, p. 395). 2 LINKING RESOURCES AND SECESSIONISM 5 that estimate the effect of economic indicators on the governing party’s chances to get re-elected (see, e.g., Brender & Drazen, 2008; Jordahl, 2006) reveal that voters especially react retrospectively to changes in macroeconomic indicators like the unemployment rate and inflation in the years prior to an election and thus tend to take a sociotropic perspective (Lewis-Beck & Stegmaier, 2000). When it comes to situations that alter the voter’s economic expectations about the future, for instance due to the discovery of resources or a change in their value and the expected benefits of its exploitation, it seems only natural to assume that voters also adopt a prospective view (see Kuklinski & West, 1981, p. 437). In one of the few cross- sectional analyses of secessionism, Sorens(2005) shows correlational evidence that voters do take cost- benefit-considerations into account when voting for a secessionist party.5

From Alesina & Spolaore(1997) to recent contributions like Boffa et al. (2016), a large literature in economics has both theoretically and empirically evaluated the reasons behind regional integration and disin- tegration. In the initial seminal paper, Alesina & Spolaore(1997) argue that the equilibrium number of nations is a result of the trade-off between the costs and benefits of being a member of a larger political entity. Other important papers include Goyal & Staal(2004) who highlight the role of size, location and diversity within regions, and Buchanan & Faith(1987) who argue that the secession option places an upper limit on the tax burden that a ruling majority can impose on the minority. Bordignon & Brusco(2001) consider whether federal constitutions should allow peaceful secession options. In an alternative model, Bolton & Roland(1997) concentrate on the fact that people in different regions might exhibit different preferences on income redistribution within the chosen political entity. In contrast to their paper, we are mostly interested in differences in resource allocation and distribution amongst regions, though we also highlight the role of economic considerations. In the following, we outline a simple model to help us to understand how economic arguments centered around the distribution of regional resources can explain the electoral success (or failure) of secessionist parties.

Assume that the utility of the representative citizen of a region r in a union of R regions contains costs and beneﬁts of integration in the union (nation state). The level of integration is indicated by Ir ∈ [0, 1], with 1 indicating full integration into the nation state and 0 indicating complete separation. As we are interested in secession, we focus on the cases where Ir = 0 and Ir = 1. As in the framework of Alesina & Spolaore

(1997), citizens bear heterogeneity costs Cr of integrating into larger units. These costs are deﬁned as:

Cr = ℎr (Ir ) (1)

5 Note that “Secessionism is a [...] demand for formal withdrawal from a central political authority by a member unit or units on the basis of a claim to independent sovereign status” (Wood, 1981, p. 110). Parties advocating secession are also often referred to as regionalist, separatist or autonomist. In our deﬁnition, separatism is the pursuit of more independence from the central state which might but does not need to culminate in secession. We keep the diﬀerences in mind, but mostly use all terms interchangeably throughout the paper. 2 LINKING RESOURCES AND SECESSIONISM 6

Costs from heterogeneity exist due to a deviation from r ’s preferences from the rest of the nation. The term refers to preference heterogeneity within the nation and can be expressed as the Euclidean norm:

|| − || · ℎr = pr pN¯ Ir , where pr is a vector representing the preferences and values of the average citizen in region r . pN¯ can be understood as representing either the preferences of the majority in the nation, the pivotal region in bargaining processes, or the median voter. Hence, heterogeneity costs increase with a larger diﬀerence between preferences in region r and the national majority (ℎN¯ ). These costs are often forgotten in the political dis- cussion because they are non-monetary and non-quantiﬁable as they relate to regional sentiments, common values, cultures, and norms Alesina et al. (1995). Secession can provide the advantage of, for instance, the provision of public goods (Alesina & Spolaore, 1997) and redistribution (Bolton & Roland, 1997) more in line with regional preferences.

Citizens also derive benefits Br from national integration from, e.g., trade, other economies of scale or enhanced public good provision. These benefits are defined as:

PR = Vi B = b (I ) + i 1 , (2) r r r R

0 with Vi = xi × pi and b (Ir ) ≥ 0. Large jurisdictions generate economies of scale in shared public institutions and infrastructure, such as administration or national defence (Alesina et al., 1995). A large, diversified economy and its respective welfare state also offers better protection against economic shocks and a more efficient provision of public goods since, as for any non-rival public good, the per capita costs are lower (Alesina & Spolaore, 1997, pp. 1028-1029). Another relevant factor in the case of a secession is the potential loss of trade advantages which exist within large jurisdictions (Alesina & Spolaore, 1997; Young, 2014).6

The value of regional resources Vi is simply modeled as the quantity of resources available xi times the price of the resource pi. Note that there might be more than one resource in reality, but this would simply mean to instead take the sum of the respective resource values. Resources can be important for the secession decision in various ways. When we think of the costs of public good provision, resource revenues can

6 In this respect, the European Union is important for secessionist movements in member states. The SNP and Plaid Cymru actively aim at making use of the political and economic advantages of the EU, which would primarily provide them with external economic security in case of a secession (Nagel, 2004). Alex Salmond, a former SNP leader, stated: “The whole debate on independence has been changed by a single idea, [...] and that’s the European Union” (see Washington Post following the link http://www.washingtonpost.com/archive/politics/2000/12/12/ eus-potential-lifts-scots-hope-of-independence/f5c65ca8-3c3d-417c-bffd-620c32225337/, last accessed on 31st October 2016). In contrast to the general EU scepticism in England, Scotland and Wales today run their own repres- entations in Brussels to promote their national self-interests (Paquin, 2002). Gehring & Schneider(2016) show that minor European states can achieve significant economic gains by making targeted use of EU key positions. 2 LINKING RESOURCES AND SECESSIONISM 7 enable regional governments to secure a similar or even higher level of public goods provision than in the case of remaining in the union. Another argument in the existing literature is that setting up and operating a complete state apparatus leads to huge direct monetary costs of uncertain extent, which might only be feasible if enough resources are available (Young, 2014; Lynch, 2003). Regarding the effect of differences in wealth levels, the political science literature has shown a correlation between the relative wealth of a region and separatist tendencies in democracies (Sorens, 2004; Van Houten, 2007). One cited reason is that wealthier regions are often the net-contributors in a union which subsidize other regions.We choose a very simple option to integrate resources in the model as there is not much additional analytical value from further complications for our purpose. It assumes that resource revenues are pooled and then evenly redistributed among all regions.7

For a given level of integration, citizens optimize in terms of consumption, leisure etc.. The indirect utility = ( , ) = ( ( ), , PR , ). for the representative citizen in region r is then given by Wr Wr Br Cr Wr br Ir Vr j,r Vj ℎr In case the region remains a part of the union or the nation state, the utility is:

PR = Vi W N = b (I ) + i 1 − ℎ (I ) (3) r r r R r r

In the case of secession, the region can keep all the beneﬁts derived from the resources. We focus on complete secession Ir = 0 with b (0) = 0 and ℎr (0) = 0, so that we get

S Wr = Vr . (4)

2.2.2 Support for Secession

S Assume citizens can vote to either stay in the nation-state (Ir = 1) and get Wr or vote for secession which N means choosing Ir = 0 and getting Wr . A citizen from region r will agree to secede and choose Ir = 0 if

S N S N Wr > Wr + σi,r ⇔ σi,r < Wr − Wr (5)

S where Wr indicates the indirect utility for the representative citizen in region R. σi,r is the citizens’ idiosyncratic strength of regional identity and follows a uniform distribution on −λr /φ and (1 − λr )/φ, with 0 < φ ≤ 1 (for details on the probabilistic voting model, see Persson & Tabellini, 2002). The parameter

7 One straightforward extension would be to integrate the usage of the resources and assume that they are spent on a public good. In the case of a true public good there might be beneﬁts from pooling resources at the national level. Still, due to the diﬀerences in preferences captured by ℎr , the level of the public good would deviate from the optimal level of region r . Another extension for a more complex model would be to consider the impact of resource wealth on the bargaining power regarding within-country redistribution. This could augment the existing literature like Dixit & Londregan(1998) and Persson & Tabellini(1996). 2 LINKING RESOURCES AND SECESSIONISM 8

λr ∈ (0, 1) can be understood as a baseline approval for secession and aﬀects the position and width of the S N distribution. In each region there is a threshold value σr = Wr − Wr for which citizens are indiﬀerent between secession and the status quo. Plugging (3) and (4) into the equation we get

PR = Vi σ = W S − W N = V − b (I ) − i 1 + ℎ (I ) r r r r r r R r r

PR R − 1 Vi ⇔ W S − W N = V + ℎ (I ) − i,r − b (I ) (6) r r R r r r R r r

The share of people π who support the secessionist party in region r is then given by πr = P r [σi,r < S N Wr − Wr ]. This yields

PR R − 1 j,r Vj π = λ + φ[ V + ℎ (I ) − − b (I )]. (7) r r R r r r R r r

We can now use comparative statics to derive the influence of regional resources and the other main factors that drive secessionism. This also helps to establish which prerequisites need to be met to find a suitable treatment and counterfactual group. A higher value for λr shifts the distribution of σi,r to the left and increases baseline approval for secession. λr can be considered the region specific strength of regionalism (see Dehdari & Gehring, 2016). Regions’ costs from preference heterogeneity can also differ strongly based, for instance, on the length of common history with the other regions and the extent and salience of factors that distinguish it from the rest. Hence, one of the main empirical challenges is to find a treatment and counterfactual which are sufficiently comparable in these respects to enable us to isolate the effect of regional resources.

π It is straightforward to see that d r > 0. If the value of resources in region r increases, secession becomes the dVr more attractive option for a larger share of people.8 The decisive question is whether changes in regional resources value really help to explain the emergence and success of secessionist parties. The assumption we make in transitioning from the model to the empirical application is that higher support for secession translates into higher vote shares for secessionist parties. This is supported by Jolly(2014) and Sorens (2005).

8 Note that a potential alternatives to secession would be decentralization and more autonomy for the region which experiences an increase in resource value. For such adaptions within federal systems it is extremely hard to reach an agreement, especially if all regions are required to agree, and it usually takes many years for a change to be implemented. A model of referendums on secession and their eﬀect on bargaining about transfer payments is provided by Eerola et al. (2004). Another question that we do not further consider are the ﬁxed costs of seceding. 3 EMPIRICAL RELEVANCE 9

3 Empirical Relevance

3.1 Anecdotal Evidence from Around the World

Secessionist movements naturally depend on a combination of multiple factors such as high group identification and the ability to solve collective action problems to be successful (Hechter, 1992, p. 269). In addition to having a geographically concentrated culturally homogeneous social group (Wood, 1981, p. 112), rational economic arguments can be pivotal. Authors like Casertano(2013), Collier & Hoeffler (2006), and Sorens(2005) even argue that group identification is sometimes only artificially created : “Eth- nic identity can provide a sense of separateness, but voters consider this separateness relevant only when it can be mobilized to achieve political and economic goals that are important to them” (Sorens, 2005, p. 307). In a cross-sectional analysis, Sorens(2005) shows that identity variables like a region-specific minority language or a recent history of independence matter, but that economic variables account for the larger part of secessionist party support and particularly for its variation.

This section provides an overview over various cases of regions where increases in regional resources or their respective value relate to secessionist movements. The main purpose of this section is to highlight the relevance of regional resources, before we proceed with our causally identiﬁed main estimation. It also demonstrates that the mechanisms are relevant for various kinds of regional resources, including human capital and land endowments (the Appendix provides further details and sources). Democratic secessionist movements that we observe around the world can be broadly grouped into three categories. First, there are of course movements where economic arguments do not play a role or at least only a very minor one. For instance, while several parties and organizations promote independence in Kashmir in India, it is ethnic motives and protest against suppression that seem to be driving these movements which also tend to be more violent in nature. While cases falling in this category are rather infrequent in democratic countries, it is important to remember that this is in line with our model: secessionist tendencies can arise in the absence of any regional resource diﬀerences.

The second category contains regions where economic arguments play a major role for the separatist discourse and are widely used in party campaigns, but where the value of regional resources does not vary enough over time to clearly distinguish their inﬂuence from other factors. Useful examples include North- ern Italy, Silesia in Poland, and Istria in Croatia. Consider the main separatist (formerly secessionist) party in Italy, the Lega Nord (LN ). One of its central political goals is more autonomy (and even independence until 2006) for the North of Italy, which possesses higher human and physical capital compared to other regions in Italy. Accordingly, it is the most developed and productive area in Italy. The tax revenues gener- 3 EMPIRICAL RELEVANCE 10 ated by superior regional resources are redistributed within Italy via a system of transfer payments. Protests against these transfers were a major reason for the call for a secession referendum in the 1990s.9 Another interesting fact is that the movement “is not based in an area that has historic claims to nationhood. Instead, the LN has attempted to invent an ethnicity for the North of Italy in order to justify its political claims for the protection of the economic interests of the region” (Cento Bull & Gilbert, 2001, p. 446).

Another example is Silesia in Poland. The Silesian Autonomy Movement exploits the fact that the region possesses extensive coal and lignite deposits and is rich in zinc, lead, and iron. The movement’s homepage states that the revenues from these regional resources should “[...] remain at our disposal. The inhabitants of the land will decide on the distribution of these funds.”10 In Croatia, the region Istria is endowed with beautiful beaches as well a ﬂourishing processing and shipping industry. The Istrian Democratic Assembly and the separatist Istrian Democratic Forum instrumentalize the redistribution of the money generated from these resources to foster their electoral success. Observers agree that “Istrian regionalism was not motivated primarily by ethnic or national belonging. Such movements [...] are driven more by economic concerns than by cultural or ethnic criteria” (Ashbrook & Ashbrook, 2008, p. 151). Other cases where economic arguments clearly contributed to secessionist parties’ behavior are the Pro Santa Cruz Committee in the Santa Cruz region in Bolivia which possesses the second largest natural gas deposit in South America, and the Republika Srpska Movement in Bosnia-Herzegovina, which is rich in minerals reaching from bauxite, to marble, and silica sand. People in the region Baluchistan in Pakistan complain that their resources “including coal as well as gas, have been exploited by the central government without adequate compensation to the province.”11

These examples highlight the importance of economic concerns for democratic secessionism and suggest that the relevant type of resources does not only comprise of natural resources. The third category of regions feature more variation in regional resource value over time: in line with our theory, these regions exhibit a positive correlation between secessionist party success and the value of regional resources. Con- sider New Caledonia ﬁrst: the former colony is a part of France, but since the 1980s features several parties like the Kanak and Socialist National Liberation Front which promote independence. The rising popularity of these parties has developed along with the rise of New Caledonia, which holds roughly a quarter of the world’s nickel reserves, ranking 5th among the nickel-producing countries worldwide. A joint venture with a Canadian multinational to exploit more of their resources is regarded by many Kanak as making separation more feasible and economically attractive. In particular, the Koniambo Project, a nickel mine in which

9 See, e.g., The Economist from 27th May 1997 at http://www.economist.com/node/150513, last accessed 31st October 2016. 10 See the FAQ section on the webpage of the Silesian Autonomy Movement at http://www.autonomia.pl/faq/, last accessed 31st October 2016. 11 See The Economist from 5th May 2005 at http://www.economist.com/node/3941524?story_id=3941, last accessed 31st October 2016. 3 EMPIRICAL RELEVANCE 11 the company invested $6 billion, has further instilled hopes for independence in the local population. As one observer puts it “resource sovereignty in New Caledonia has come to be seen by independence leaders as a path to political independence” (Horowitz, 2004, p. 287).

The French-speaking Canadian region Québec clearly differs from the rest of Canada linguistically and also to some degree culturally. It features the Parti Québécois and the Bloc Québecois which represent separatist interests on the national and state level. It is is an interesting case, as the cultural differences are mostly constant over time, whereas the parties’ success varies strongly. Historically, Québec was rich in natural resources like gold, iron, copper and wood. Accordingly, a central party claim revolved around the fact that the Québecois transfer more than C$53 billion per year to the federal redistributive system. The electoral success of the separatist parties diminished over the years, however, as Québec’s economy only managed to grow significantly less than the rest of Canada. Simultaneously, we observe a decline in the parties’ shares in the national (from 13.5 percent to 4.7 percent) and the regional parliament (from 49.3 percent to 19.3 percent) over the 1993 to 2015 period (see Appendix A, Figure 2).

Other cases also provide plausible exogenous within-country variation. Greenland enjoys far-reaching autonomy but remains a part of Denmark in several respects including justice, defense, and foreign affairs. In Greenland’s parliament, the Inuit Ataqatigiit and the Forward Party represent the struggle for more autonomy or full independence from Denmark. As Greenland’s mostly fishing-based economy was stagnant for a long time period and almost half of public spending was financed by grants from Denmark, their electoral success was limited. The discovery of oil and the fact that, due to the melting of the Arctic ice, larger areas become feasible for mining (rare metals and radioactive substances), the secessionist parties gained support at the expense of the largely Danish Democrat Party in the 2009 elections.12 In 2008, a non-binding referendum on more self-governance won in a landslide with 21,355 votes to 6,663.13 However, the drastic collapse in crude oil prices since 2015 has made most Arctic oil unprofitable to exploit and led “Greenland to put off plans to split from Denmark”.14

All cases in this second category indicate a positive relationship using variation over time within the same country. Ideally, however, we would ﬁnd a suitable counterfactual region within the same country. Two examples that partly fulﬁll the criteria are Galicia and Catalonia in Spain, and Flanders and Wallonia in Belgium. Galicia and Catalonia both feature separatist and secessionist parties: the Galician Nationalist Bloc

12 See The Economist from 15th July 2012 at http://www.economist.com/blogs/graphicdetail/2012/06/ daily-chart-9 and from 31st March 2013 at http://www.economist.com/blogs/newsbook/2013/03/ economist-explains-why-greenland-election-global-implications, last accessed 31st October 2016. 13 See The New York Times from 26th November 2008 at http://www.nytimes.com/2008/11/27/world/europe/ 27greenland.html?_r=0, last accessed 31st October 2016. 14 See The Economist from 21st January 2015 at http://www.economist.com/news/europe/ 21640224-falling-crude-prices-are-forcing-greenland-put-plans-split-denmark-independence-ice, last accessed 31st October 2016. 3 EMPIRICAL RELEVANCE 12 and Compromiso por Galicia in Galicia, and the Republican Left of Catalonia, the Democratic Convergence of Catalonia, as well as the Popular Unity Candidacy in Catalonia. Both regions’ preferences differ significantly from the rest of Spain as they possess, for instance, a distinct history and a language which linguists describe as unintelligible to Castilian Spanish (Lewis et al., 2009). However, the success of secessionist parties differs drastically between the two regions. While newspapers are filled with articles covering secessionist party success in Catalonia, the Galician regionalist parties jointly receive not more than 20 percent of the votes. The most obvious difference between Catalonia and Galicia is that Catalonia has the highest regional GDP in Spain, based on a strong industrial base and superior human capital endowment, exemplified by its leading position in high-tech exports. According to the Democratic Convergence of Catalonia’s manifesto, Catalonia contributed 19.5 percent to the total resources of the central state, but received only 14 percent of the spending. The manifesto of the Republican Left complains about the usage of Catalan resources to finance public goods like high-speed trains, airports, and defense, and highlights the potential benefits of being able to manage and keep Catalonia’s resources to themselves. As the Financial Times reports, a majority of Catalans feels “Madrid takes too much of local income to redistribute elsewhere.”15

The Belgian case comes closer to our ideal scenario. Basically, Belgium comprises two ethnically and culturally distinct regions: a French speaking part, Wallonia, and a Flemish (Dutch) speaking part, Flanders. There is extensive variation in the relative value of their respective regional resources over time, which we can observe along with the strength of secessionism and secessionist parties. Up to the 1960s, Wallonia was one of the richest regions in Europe due to natural resources like coal and a comparative advantage in the leading sectors at that time (such as coal mining, steel making, and related activities; see Mnookin & Verbeke, 2009). While there were Flemish independence movements building on the suppression of the Flemish language and the political dominance of the smaller French part since the foundation of Belgium in 1830, support for secessionism never really took oﬀ until the economic situation reversed. Declining demand for coal and steel on the one side, and modernization and the increased value of possessing the important port of Antwerp on the other side made Flanders’ regional resources relatively more valuable compared to Wallonia’s. As our model would predict, this reversal of fortunes clearly correlates strongly with the increasing vote share for secessionist parties and general support for secession in Flanders. In 2012, the secessionist New Flemish Alliance even became the largest party in the Belgian federal elections. The party argues that “wealthy Flanders should not be subsidizing poorer Wallonia, whose regional government is alleged to be wasting money. Flemish nationalists feel strongly that their region is not receiving its fair share of the revenues that it contributes to the national economy.”16

15 See the Financial Times from 26th September 2012 at https://www.ft.com/content/ bad90798-07f4-11e2-9df2-00144feabdc0, last accessed on 31st October 2016. 16 See http://knowledge.wharton.upenn.edu/article/secession-answer-case-catalonia-flanders-scotland/, last accessed 31st October 2016. 3 EMPIRICAL RELEVANCE 13

The abundance of anecdotal, correlational evidence further convinces us in our assessment that economic considerations indeed play an important role in explaining the success of democratic secessionism, and that regional resources in various categories and the distribution of the associated revenues are the key economic factors. The listed cases have demonstrated that richer regions feature stronger secessionist parties, that changes in resource value over time correlate with the success of these parties, and, as Spain and Belgium demonstrate, that comparable regions exhibit drastically diﬀerent support for secessionism based on the value of regional resources. We now turn to Wales and Scotland in the United Kingdom to identify whether the described pattern is indeed a causal relationship. As we argue, this case provides a natural experiment that we can exploit to compare a treatment with a suitable counterfactual group.

3.2 Scottish Versus Welsh Independence and the Discovery of North Sea Oil

The Scottish “nationalist discourse traditionally has had a weak cultural dimension. Instead, nationalist claims are based on rather practical arguments about institutions, policy and accountability” (Holitscher & Suter, 1999, p. 272).17 Out of economic and political weakness, Scotland opted for the union with England in 1707 in order to belong to (what was) the strongest and most influential nation in the world at that time (Bryant, 2006, p. 80). Although the majority of the Scots have not dominantly developed strong ’British’ national sentiments since then (Bryant, 2006, p. 5), there was no need to pursue a secessionist strategy due to the economic and military strength of the Empire (Bryant, 2006, p. 62-65; Gourevitch, 1979). The SNP, established in the 1930s, had little political relevance before and in the first years after the Second World War. Similarly in Wales, the Plaid Cymru, founded in 1925, has self-government and secession on its political agenda, but failed to gain much ground beyond narrowly defined supporter groups.

Things began to change in the 1960s. Owing to the peaceful political developments in Europe and the decline of the British economy, safety issues were increasingly replaced by economic concerns (Holitscher & Suter, 1999; Paquin, 2002). Neither the unloved Tories nor the Labour governments from 1964 onwards managed to improve the economic situation (Gourevitch, 1979; Paquin, 2002). As a consequence, the SNP and similarly Plaid Cymru in Wales gained their ﬁrst small electoral successes in by-elections by the end of the 1960s.

It was around that time when the first oil was discovered in the North Sea.18 After the Geneva Convention (1958) confirmed the nations’ coastal rights, and offshore gas was discovered by the Netherlands in the

17 For details on the historical and cultural background, see Bryant(2006), McCrone(1992), and Paquin(2002). Wales has already been in the union with England by annexation since the 16th century (Bryant, 2006). 18 As Scotland and England both have a significant amount of natural gas (about 50 percent of the UK gas are in Scottish waters; see Kemp & Stephen, 2000, pp. 9-12) and the value of oil is by far higher (Brocklehurst, 2013), the discussions and campaigns have always explicitly been about the oil on which we will also concentrate in our analysis. 3 EMPIRICAL RELEVANCE 14 early 1960s, “[t]he hunt was on for North Sea oil” (MacKay & Mackay, 1975, p. 184). After several years of unsuccessful exploratory drillings in the British Sector, “oil companies were becoming disillusioned with the prospects of finding oil in the North Sea” and commonly gave up (Whaley, 2010, p. 77). Against this background, the discovery of the Forties oil field off the Scottish coast appears even more like an exogenous shock, upon which further giant discoveries would follow.19 A summary table containing all (giant) discoveries in the Scottish sector is provided in Appendix B.

Figure 1: Oil Fields The map shows the hypothetical Scottish maritime boundaries as well as the locations of the oil ﬁelds discovered in the UK sector. Oil ﬁelds are displayed in the map by using ArcGIS.

To calculate the Scottish share of British oil, a maritime border which is equidistant in all points to the Scottish and English coast, being also used to deﬁne Scottish ﬁshery grounds, is generally applied (UK

19 See, e.g., The Guardian following the link http://www.theguardian.com/business/2014/oct/12/ forties-oil-field-50-north-sea-uk-offshore-bp, last accessed on 31st October 2016: “BP discovered gas in the North Sea in 1964, but it was not until 1970 that it [...] had found signs of hydrocarbons more than 3,000 metres below the seabed around 110 miles from Aberdeen.” 3 EMPIRICAL RELEVANCE 15

Statistics Authority, 2013, pp. 6-7).20 There is no doubt that an independent Scotland would own the mineral rights for the mentioned sector (MacKay & Mackay, 1975, p. 168). Building on this foundation, the hypothetical share of oil production in Scottish waters has generally exceeded 90 percent of the total UK oil production (see Kemp & Stephen, 2000, and ﬁgures in Appendix E). The monetary revenue of North Sea oil is mainly represented by government tax receipts, which make up about 75 percent of the gross revenue (Kemp et al., 1983, pp. 120-125). “In the early 1980’s when oil prices peaked the share exceeded 98 percent” (Kemp & Stephen, 2000, p. 8).21 The known remaining oil reserves are estimated with a current monetary value of around £120 billion (UK Statistics Authority, 2013, p. 32). In recent years, oil revenues would represent 10-20 percent of the total tax revenue in an independent Scotland (UK Statistics Authority, 2013, p. 31).

For the most part, the direct economic impact of the oil sector was concentrated in Scotland (Scott, 2004, p. 339). It is estimated that up to 90,000 new jobs were created (Lee, 1995; MacKay & Mackay, 1975, p. 136), particularly in the rural areas of Grampian and Highlands as well as Aberdeen (Lee, 1995). The country proﬁts from a large amount of money invested in the oil sector and its multiplier eﬀects (Kemp & Stephen, 2000, p. 1), which lead to increased income and GDP. Due to the high amount of oil-related exports, an independent Scotland is supposed to have a strong balance of payments (Kemp et al., 1983; MacKay & Mackay, 1975, p.173).

The SNP had been very successful in making political capital out of the situation in the early 1970s.22 Their distinctive “It’s Scotland’s oil” campaign, their bold and simple per capita calculations of oil revenue (Lee, 1976, pp. 307-314), and the claims the Westminster politicians would sell out ‘Scottish oil’ (Lee, 1976, p. 312) by exploiting the ﬁelds too rapidly (MacKay & Mackay, 1975, pp. 24-30) struck a nerve in Scotland. Consequentially, the party’s rise has often been related to the oil: Most prominently, Gordon Brown, the later British Prime Minister suggested in an essay in 1975 that “the rise of modern Scottish nationalism is less an assertion of Scotland’s permanence as a nation than a response to Scotland’s uneven development [...] and their (oil-ﬁred) expectations at a Scottish level” (Brown, 1975). Indeed, it is tenable to assume that

20 Other plausible alternative borders would presumably not change the results much as “there are just a handful of fields, and not very important ones now” between the imaginable lines (Brocklehurst, 2013). However, our choice is the most conservative as the alternative would additionally localize further southern oil fields in the Scottish sector. More details are provided in Appendix E. 21 An official report for the 2014 referendum by the (UK Statistics Authority, 2013, p. 31) reveals that the UK Treasury’s tax revenues from the Scottish oil sector have been volatile, being “as high as £27 billion in 1984-85, as low as £1.5 billion in 1991- 2 and most recently at £11.2 billion in 2011-12 in current prices.” Opponents of independence frequently use the described volatility as an argument against Scotland’s subsistence (Lynch, 2003). 22 According to Lynch(2003), the SNP already began to explore the potential impact of an oil discovery in years before oil actually was discovered: “The party’s research officer had studied the economic benefits of oil development in Canada [...]. The SNP gathered a considerable amount of statistical data in the field of oil and oil development, [...] superior to the UK government and some of the oil companies themselves. SNP estimates and projections of oil deposits in the North Sea, along with future oil revenues and economic development associated with oil, became regular features of media coverage, especially as the SNP focused on the economics of independence” (Lynch, 2003, p. 5). 4 DATA AND EMPIRICAL STRATEGY 16 the locally concentrated oil discoveries help the SNP to create a picture of an independent Scotland being among the richest nations in the world (Collier & Hoeffler, 2006; Jolly, 2007) . Hence, the main hypothesis to be tested in our analysis is whether the electorate would increasingly vote for the national party after oil fields have been discovered. Lynch(2003, p. 6) concludes that “for the SNP, nothing has ever replaced oil in its economic case for independence.”23 The nature of resource discoveries possibly could have strengthened the SNP’s momentum even more:

“[Natural resources] are usually ‘discovered’ at a particular moment, [...] perhaps being a gold rush. [T]he prices of natural resources are volatile, with periodic spikes, so that there are precise moments when the economic value of a particular resource becomes dramatically valuable.” Collier & Hoeﬄer(2006, p. 5)

Led by this argument, we also expect a stronger inﬂuence of oil discoveries on nationalist party support when the oil price – and simultaneously the current value of the discovered oil reserves – is higher. Before testing the hypotheses in section5, we describe the dataset used for analysis in the subsequent section, starting with the panel structure of the data.

4 Data and Empirical Strategy

4.1 Data

There is no doubt that an independent Scotland would own the mineral rights for the sector mentioned in the previous section (e.g., MacKay & Mackay, 1975, p. 168). Building on this foundation, we analyze the electoral outcomes for 72 Scottish and 40 Welsh constituencies in British general elections (GEs) over the 1945-2001 period. There are several reasons for choosing this time period.24 First, we are able to rely on such a long period since we managed to collect election results on the constituency level for both Scotland and Wales back to 1945. This helps us to reliably establish whether the parallel trends assumption is indeed justiﬁed. Second, the long period is informative as the respective nationalist parties in both countries – SNP and Plaid Cymru – were clearly the most prominent organised groups supporting independence during the whole period and did not experience uniﬁcations or split-ups. In addition, both have in common that they played only a minor role in Westminster politics until the 1960s (Paquin, 2002).

23 There are also few authors like Miller et al. (1977, p. 226) and Miller(1981) who rate the inﬂuence of oil not so high and deem it rather as a supportive factor than the trigger for the party’s success, arguing that the SNP’s electoral success began with a by-election in 1967 (Mitchell et al., 2012, p. 61). 24 We do not use local elections: massive changes in local government organization in the early 1970s make the pre- and post- treatment period unreliable comparisons. Generally, both SNP and Plaid Cymru have always performed better in local elections (see McGuinness et al., 2012; Sorens, 2004) where other factors dominate election campaigns than in GEs (Sorens, 2005). 4 DATA AND EMPIRICAL STRATEGY 17

Third, we did not include election results after the 2001 GE because – as a consequence of the implement- ation of further local government competencies – the number of constituencies changed and Scotland lost 13 of its previous 72 seats in the House of Commons (McGuinness et al., 2012, p. 11). Our panel design is only identical to a DiD approach, and enables us to draw causal inferences, with the consistent inclusion of time and unit fixed effects. Note that the period from 2001 onwards coincides with both rising oil prices and increases in the vote share of the SNP; including it would thus most likely strengthen our results.25 Constituency boundaries have also been changed several times within our sample period.26 From 1945 to 2001, the amendments of five redistribution rounds have come into force: in 1947, 1954, 1969, 1983, and 1995. As a result, the number of constituencies ranges between 71 and 72 for Scotland and 35 and 40 for Wales. We resolve this issue by projecting the election results to the constituency boundaries in the 2001 GE, on which we draw on for our estimations. The detailed algorithm used to match the constituencies and results is described in Appendix C.

The ﬁnal dataset comprises data on 16 GEs held in the UK since 1945, which were collected from Brancati (2015) and Outlaw(2012). 27 We also include 91 by-elections which were held in either Scottish or Welsh constituencies when an incumbent had to be replaced (e.g., due to death or resignation).28 The dependent variable Nationalist vote share is deﬁned as the percentage share of votes received by the SNP or Plaid Cymru in UK GEs and by-elections in a constituency i at time t. If no nationalist candidate stands for election, Nationalist vote share is coded as 0.29

Figure2 displays the shares of Plaid Cymru (in light red) and the SNP (in dark blue). Both parties received only a small share of votes in the ﬁrst years of the observation period. Single by-election victories in 1966 (Plaid Cymru) and 1967 (SNP) marked the beginning of the parties’ ascent. Subsequently, the SNP was more successful, receiving more than 30 percent of the Scottish vote in October 1974; Plaid Cymru’s culmination point was 14.3 percent in the 2001 GE (McGuinness et al., 2012). The time interval in our

25 Political Scientists also argue that voting patterns changed after 2001 as the increase in competencies for the regional governments decreased the relative importance of British GEs (Jones & Scully, 2006, p. 129). If this would aﬀect Scotland diﬀerently than Wales (as a time varying variable which is not orthogonal to the treatment), it could contaminate our results. 26 Constituency boundaries are reviewed by Boundary Commissions in each of the four UK regions (see http://www. parliament.uk/about/how/elections-and-voting/constituencies/, last accessed on 31st October 2016). These commissions publish reports on the constituencies and recommend changes to the parliament, if deemed necessary. In order to maintain continuity of constituencies as well as equal parliamentary representation in terms of population, the commissions are supposed to recommend changes only if unequal population distribution exceeds a certain threshold. For details about the redistribution process, see http://aceproject.org/ace-en/topics/bd/bdy/bdy_gb, last accessed on 31st October 2016. 27 GEs have been held in 1945, 1950, 1951, 1955, 1959, 1964, 1966, 1970, 1974 (February), 1974 (October), 1979, 1983, 1987, 1992, 1997, and 2001 McGuinness et al. (2012). 28 Data for by-election results are provided by Pippa Norris following the link https://sites.google.com/site/ pippanorris3/research/data, last accessed on 31st October 2016. For the importance of by-elections see, e.g., Miller et al. (1977); McGuinness et al. (2012); Harvie(1995); Mitchell et al. (2012). 29 Miller et al. (1977) show that independence is the main criterion to vote for the nationalist parties. Hence, the share of votes for a nationalist party is commonly taken as a proxy for the public support for independence (Sorens, 2005). Brand et al. (1994) notes that the amount of protest voters can be neglected. 4 DATA AND EMPIRICAL STRATEGY 18 study is biannually, but some of the control or treatment variables are only available at a yearly frequency. Note that if only one by-election was held in a particular half-year, this observation is fully captured by the biannual-FE.

Figure 2: Election Results Share of votes received by SNP/Plaid Cymru in GEs and Westminster by-elections. The graphic displays constituency results and the total average vote share of the two parties in GEs.

We collected data on oil discoveries mainly from UK government websites.30 We used GIS software as well as various other sources to verify the size and date of each individual discovery, which was reported there. Sources are listed in the Appendix. With regards to oil discoveries, we distinguish between giant oil fields, which includes all fields above 500 million stock tank barrels (MMstb.), and all oil fields, which includes all fields above 50 MMstb. Discovery (giant) and Discovery (all) indicate the number of giant (all) oil fields that have been discovered in year t. An oil field is classified as ‘giant’ if it contains estimated ultimate recoveries of 500 million barrels of oil or more before extraction begins (Ivanhoe & Leckie, 1993). 31 We make use of this distinction as Lei & Michaels(2014) argue in their study on the influence of oil on internal armed conflicts that ‘giants’ provide the most precise measurement as they function as a clear signal of massive windfall profits. This notion is supported by Arezki et al. (2015), who emphasize that, in contrast to small oil discoveries which do not contain much new information, ‘giants’ “signal significant increases in production possibilities in the future” (Arezki et al., 2015, p. 17). In our framework it seems plausible that

30 A main data source for the oil discoveries was https://www.gov.uk/guidance/oil-and-gas-uk-field-data# uk-oil-and-gas-reserves-and-resources, last accessed on 31st October 2016. 31 Worldwide ‘giants’ are estimated to account for 74 percent of the estimated global oil reserves although less than 1 percent of all oil ﬁelds are ‘giants’ (Ivanhoe & Leckie, 1993). 4 DATA AND EMPIRICAL STRATEGY 19 oil discoveries need to exceed a certain threshold size until they are realized by voters or provide enough potential to be instrumentalized by the nationalist party. In addition to the number of discoveries, we also collected and coded the amount of oil discovered within a year t (Amount of new reserves). The variable Oil price is the real price of Brent crude oil in constant 2001-US-$, which is the major benchmark for oil produced in the North Sea. Oil price is coded as the annual average of the Brent price.32

Deviating non-linear trends in major macro variables between Scotland and Wales could bias our results. Thus, we control for the most important macro developments, for which we could ﬁnd data over the whole sample period. Unemployment rate indicates the rate of registered unemployed (“Claimant Count”) for Scotland and Wales, respectively, as a yearly average.33 In opinion polls, voters regularly mention unemployment as the most important economic issue inﬂuencing their voting decision (Zirakzadeh, 1989, p. 333). Following Scott(2004), we can also interpret it as a rough proxy for public welfare, as it correlates with other indicators like “incomes, the quality of jobs available, the proportion of lone-parent families, crime, health and housing conditions” (Scott, 2004, p. 333). In a similar vein, Regional GDP per capita, which measures the Scottish/Welsh GDP per capita as a percentage share of the UK average, proved to be an important factor as well (see section 2.2).34 Another potentially interesting control variable would have been relative government expenditure, but this is only accessible for the period after 1979 (see Heald, 1994).35

However, the literature does not suggest that the complex topic of fiscal redistribution is dominating campaigns in Scotland and Wales directly. Lynch(2003) states that these issues are “too obscure for the public to understand or become particularly concerned with. [...] Basing a campaign for independence around abstract fiscal issues at a time of growing public spending is not a strategy likely to succeed in the short- term”. Based on Sorens(2005), we picture relative GDP as a proxy for regional tax payments indicating the relative economic strength of a region. Obviously, as we estimate the DiD-strategy in a panel framework, we will include country and time-fixed effects. We will also use country-specific linear time trends, as well

32 The oil price is taken from Baumeister & Peersman(2013) who provide a monthly series for Brent, dating back to 1957. For previous years, the Brent price is projected using the price for the sort WTI. These data are provided by the Federal Reserve Bank of St. Louis (https://research.stlouisfed.org/fred2/series/OILPRICE/, last accessed on 31st October 2016.). Details on the coding of the variables are provided in the Appendix. 33 Data for Unemployment rate are retrieved from Mitchell(1988), for all years prior to 1965, and the Office for National Statistics (ONS) following the link http://www.ons.gov.uk/ons/rel/lms/labour-market-statistics/june-2015/ dataset--labour-market-statistics.html, last accessed on 31st October 2016, for the subsequent years. 34 Data sources: Regional Accounts (Office for National Statistics), for 1971-2001; Scott(2004, p. 338), for 1951, 1954, 1958, 1962, 1966; Lee(1995, pp. 53-57), for 1945-1970 (for Scotland; approximately from graphic); missing values for Wales in the 1940s, 1950s, and 1960s are filled using a linear approximation. Values for Wales before 1951 are assumed as constant. 35 Lee(1995, p. 140) remarks that “it is clearly not possible to compute an exact net surplus or loss” for the regions. Yet, estimates suggest that, even after the introduction of the so-called Barnett-Formula in 1978 – a determined mechanism that was supposed to yield cohesion of central government expenditure and was installed in order to prevent internal disputes on this topic (Bryant, 2006, p. 54) – Scotland still receives more expenditure than other regions (Heald, 1994, p. 157). See Bryant (2006, pp. 54-55) and Heald(1994) for details about the impact and amendments of the Barnett-Formula. 4 DATA AND EMPIRICAL STRATEGY 20

Table 1: Descriptive Statistics N Mean SD Min Max Nationalist vote share 1883 10.25 11.89 0 67.05 Discoveries (giant) 1883 0.48 0.98 0 3.00 Discoveries (all) 1883 1.40 2.26 0 7.00 Amount of new reserves 1883 1.19 2.04 0 5.67 Scotland 1883 0.65 0.48 0 1 Oil price 1883 25.83 19.10 7.62 81.39 Unemployment rate 1883 4.95 3.11 1.80 13.10 GDP per capita 1883 90.58 5.21 78.50 102.40 Near border (50) 1883 0.25 0.43 0 1 Near border (75) 1883 0.40 0.49 0 1 Near border (100) 1883 0.58 0.49 0 1 Coastal access 1883 0.57 0.49 0 1 Distance to Aberdeen 1883 311.11 207.57 3.19 641.06 Avg. soil suitability 1883 3.76 1.26 0.17 5.32 Ruggedness index 1883 53.37 36.72 1.93 170.47 The table shows descriptive statictics for all variables used in the analysis over the 1945-2001 period. N = number of observations, Mean = arithmetic mean, SD = standard deviation, Min = minimum value, Max = maximum value. Reserves are measured in million stock tank barrels (MMstb.), distances in km. For details on the other variables see the Ap- pendix. as constituency-specific time trends and fixed effects. Adding below-national level covariates should mostly result in more efficient estimation. Descriptive statistics are provided in Table1.

4.2 Empirical Strategy

We argue that we can apply a difference-in-differences (DiD) approach with Scotland as the treated group and Wales as the control group to estimate the causal effect of oil discoveries on the vote share of nationalist parties. This comparison is inspired by Collier & Hoeffler(2006), who briefly discuss that Wales would form a suitable counterfactual. The need to rely on DiD arises from potential omitted variable bias when using only over-time variation within Scotland, as the treatment can obviously coincide with other time- specific events. DiD enables a causal estimation based on the assumption of common trends between treated units and counterfactuals, which we argue is plausible when examining the situation in detail. First, prior to the earliest discoveries, oil played as little a role for the SNP as for the Welsh Plaid Cymru. This is supported by historical evidence which shows that no one expected large discoveries prior to 1969. Myles Bowen from Shell is quoted as saying: “in May 1969 the view was that all the worthwhile gas fields in the Southern North Sea had been found, while the search for oil in the north was doomed to failure” (Whaley, 2010, p. 77). The initial discovery can thus be regarded as an exogenous shock and natural experiment, 4 DATA AND EMPIRICAL STRATEGY 21 which we exploit to test our hypotheses. We do not rely only on the distinction between the pre- and post discovery period, but use the number and extent of discoveries over time in a panel framework. Note that the probability of finding a new oil field could correlate with previous finds in areas nearby and to some degree on the intensity of exploration (Lei & Michaels, 2014; MacKay & Mackay, 1975). Yet, the degree of uncertainty is high and in particular giant discoveries are scarce, which is why along with most authors we treat the individual discovery and discovery year as exogenous (Arezki et al., 2015; Lei & Michaels, 2014).

Second, there is no doubt that an independent Scotland would own the mineral rights of the aforementioned sector (Kemp & Stephen, 2000; MacKay & Mackay, 1975). A document by the UK Statistics Authority (2013, p. 31) reveals that the UK Treasury’s tax revenues from the Scottish oil sector have been as high as £27 billion in 1984/85. These locally concentrated oil resources helped the SNP to create a picture of an independent Scotland being among the richest nations in the world (Jolly, 2007, p. 123), as the Scots’ per capita oil revenues “would rise ten-fold if the other 45 million people were excluded” Collier & Hoeﬄer (2006, p. 7). As described in section 3.2, the overriding majority of British oil has been discovered in what can be described as the Scottish North Sea sector (see Figure 1).36 The Welsh voters’ preferences for independence should thus not be aﬀected by these discoveries.

Third, both countries have independence movements manifested in nationalist parties (Jolly, 2007, p. 121). Thus, an observable and comparable dependent variable Nationalist vote share exists for both regions. SNP and Plaid Cymru are sufficiently alike to regard them as a treatment and control group facing – to a large extent – similar success in elections in absence of the treatment. The common trends assumption requires that the nationalist party’s electoral performance in Scotland (the treatment group) would not differ from the Welsh nationalist party’s counterfactual trend in absence of the treatment. Figure 3 shows the constituency level results prior to the first oil discovery in 1970, as well as their average trend separately for Wales (light red) and Scotland (dark blue). We observe nearly indistinguishable linear parallel trends. If anything, the trend of Plaid Cymru is a little more positive, which would bias against finding a significant positive effect for the SNP after the oil discoveries following 1970. Regressing a Scotland-specific linear time trend on Nationalist vote share prior to 1970 yields an insignificant coefficient (p-value = 0.699).

The inclusion of the time-varying region-specific control variables requires the trends to be parallel conditional on these observed variables. As our main results hold in both cases, it seems that the influence of these factors on treated and untreated units is orthogonal to the treatment. We include the control variables nonetheless in most specifications as they increase efficiency. As is common practice, the DiD estimation with multiple time periods also assumes that the treatment “has the same effect in every year” (Wooldridge,

36 There was one discovery oﬀ the Welsh coast in 1990. The Douglas ﬁeld is however rather small with an estimated 225 MMstb., and it seems highly unlikely that one single discovery could be exploited succesfully by Plaid Cymru. 4 DATA AND EMPIRICAL STRATEGY 22

2010, p. 151).

Besides these technical arguments for the correctness of the DiD approach and particularly for the Welsh constituencies as a control group, it is also reassuring that based on a qualitative comparison of the two parties they exhibit many similarities, particularly with regards to their motivation and goals. Both movements put emphasis on the fact that Britain is not a homogeneous nation, comprising regions which have cultural identities distinct from England (Fusaro, 1979). As the SNP and Plaid Cymru are geographically limited parties, they can be characterized as the biggest movements promoting their regional interest and full independence on a more central level (Fusaro, 1979). 37

Figure 3: Linear Pre-Trends The graphic shows the share of votes received by the SNP/Plaid Cymru in elections before 1970. The dashed lines indicate the linear trends of the two parties in the period prior to the ﬁrst oil discoveries. The unconditional linear trend of Plaid Cymru’s vote results is stronger than the trend of the SNP.

The SNP and Plaid Cymru can be classiﬁed as nationalist left-of-centre parties (Mitchell et al., 2012; Levy, 1995), usually strongly opposing what they call “English Tory government” (Levy, 1995, p. 296). Both parties promise, for instance, to strengthen the social welfare system and aim for more social equality. With regards to their regions’ connection to England, they build up a picture of ‘internal colonialism’ (Fusaro,

37 While Scottish nationalism is predominantly influenced by an economic and political calculus, regional ideology is supposedly even stronger in Welsh nationalism based on preserving Welsh culture and language (Fusaro, 1979; Lee, 1995, p. 298). This is exemplified by the stronger roots of the Welsh language, which is the dominating spoken language in some parts of the region, compared with the unusual Gaelic in Scotland, which is only spoken by 1 percent of the population (Sorens, 2005, p. 306). This is among the only significant differences between the nationalist tendencies in the two countries as the political positions and aims of the parties are quite similar (Bryant, 2006, p. 291). The cultural uniqueness of Scotland can instead be spotted in the greater conservation of original Scottish institutions, e.g., an education system differing from the rest of the UK (Fusaro, 1979). 4 DATA AND EMPIRICAL STRATEGY 23

1979; Nagel, 2004). The notion that the two parties can be treated as belonging to the same kindred group is not far-fetched: after the 1974 GE, the Plaid Cymru leader Gwynfor Evans stated in Westminster that there would now be “more nationalists in the House than Liberals” (Fusaro, 1979, p. 365).

There is no reason for assuming that Welsh voters or the Plaid Cymru would react to oil discoveries off the Welsh coast in a different way than the Scottish did. This question reflects the DiD assumption that the treatment would have the same effect on the control group as it has on the treated group, even though there might be slight differences in detail between the two parties. Two anecdotal examples underline why this assumption holds true. Firstly, Plaid Cymru has also attempted to run election campaigns on the issue of natural resources. Proclaiming the slogan ‘Hands off Welsh water’, it decried the overexploitation of Welsh springs to supply English cities Collier & Hoeffler(2006, p. 7) and also protested against rising water charges in Wales (Levy, 1995). Secondly, in the 1970s the already quoted Plaid Cymru leader Gwynfor Evans “constantly affirm[ed] that oil lies under the Celtic Sea, as if trying to wish it and Welsh independence into existence” (Lee, 1976, p. 307). This shows that Plaid Cymru recognized the potential electoral gains from arguments based on natural resources that voters in their respective region are claiming for themselves as well.

The main estimation equation reads as follows:

= α + δ × + 0 γ + λ + ϑ + ε Nationalist vote sharei,c,t Discoveries (giant)t Scotlandc Xc,i,t c/i t i,t , with Nationalist vote share being the vote share of the respective nationalist party in constituency i in country c ∈ Scotl and; W al es at time t. The treatment effect is measured as δ, the coefficient of the interaction term of Discoveries (giant) and Scotland (the binary indicator for Scotland). It captures the average treatment effect on the treated (ATT). The main effect of Discoveries (giant) is fully captured by biannual-FE ϑ. Depending on the specification, the main effect of Scotland is either captured directly by a binary variable λ or by fixed effects λ for each constituency. Using λ in the panel DiD-setting mainly serves to increase efficiency, but would also pick up any constituency-specific characteristics that are time- invariant, e.g., a specific culture or the degree of urbanity. The same holds for the two time-variant controls. X 0 is the vector containing the control variables and γ the respective coefficient vector. α is the constant and ε the error term.

We will also show results from equations like

= α + δ × + 0 γ + λ + ϑ + τ + ε Nationalist Vote Sharei,c,t Discoveries (giant)t Scotlandc,i Xi,t i t i/cT c,i,t , where T is a linear trend variable, and τi/c is a country- or constituency-specific trend coefficient. As our treatment is “randomly” switched on and off when oil is discovered, we cannot as easily test for pre- and 5 RESULTS 24 post-treatment effects as in more simple settings. Allowing for country-specific trends relaxes the common trend assumption as they allow linear country specific deviations from the trend.38 This serves as an informative test for the robustness of the results.

This approach differs from simple two-period DiD in including multiple time periods and having a continuous treatment variable Discoveries (giant) (Wooldridge, 2010). The continuous nature assumes a linear effect of discoveries, while it is possible that each additional discovery has a diminishing effect. Note that this would again bias against finding a significant treatment effect compared to using a simple dummy variable. We opt for this choice as it contains more information which we use to interpret the results. δ is thus the average treatment effect on the treated (ATT), the additional vote share that the SNP receives compared to their counterfactual trend indicated by the performance of Plaid Cymru caused by one additional oil discovery. Standard errors allow for arbitrary correlation both within constituencies and across units at one point of time (two-way clustered, see Baum et al., 2010). The results are robust to alternative clustering choices, as we discuss later. . The data cover the 1945-2001 period and include a maximum of 1883 constituency-half-year observations.

5 Results

We start by looking at the simple fixed effects regression of the SNP vote share on the number of giant oil discoveries in the year of and the year before the election, focusing only on Scotland. It includes time-fixed effects and the two time-varying control variables Unemployment rate and GDP per capita. The coefficients of both control variables are positive. Constituency-fixed effects pick up any time-invariant omitted variables, but the estimates could still be biased by time-varying omitted factors. The conditional correlation is clearly positive: the coefficient of Discoveries (giant) is 4.494 and significant at the 1-percent level. One additional giant oil discovery would thus be linked to an increase of about 4.5 percentage points for the SNP. This correlation might be biased if other developments affect the vote share of the SNP which are also correlated with oil discoveries, which is particularly concerning given the absence of a wide array of control variables at the country level.

Hence, we turn to the first DiD estimation in column 2. This adds Welsh constituencies to the equation, as well as time-fixed effects, a binary variable that takes on the value of 1 for all constituencies within Scotland, and the interaction of this variable and the oil discoveries Discoveries (giant) × Scotland. The treatment effect

38 Technically, including treatment-specific trends relaxes the common trend assumption, which refers to the change, i.e. the first derivative of the dependent variable. Instead, we need only to assume parallel growth between treated and untreated units, i.e. the same second derivative, while the first derivative is allowed to differ. 5 RESULTS 25

Table 2: Regression Results Nationalist Nationalist Nationalist Nationalist Nationalist Dependent Variable vote share vote share vote share vote share vote share Scotland × Discoveries (giant) - 3.261 2.862 1.923 1.926 [0.799] [0.744] [0.882] [0.898] Scotland - 2.263 −3.500 - - [1.762] [1.439] Discoveries (giant) 4.494 - - - - [0.253] Unemployment rate 0.977 - - 1.737 1.754 [0.099] [1.097] [1.206] GDP per capita 1.185 - - 0.725 0.721 [0.093] [0.214] [0.231] p-value: Scotland × Discoveries (giant) - 0.000 0.000 0.029 0.032 Biannual fixed effects no yes yes yes yes Constituency-fixed effects yes no no yes yes Linear time trend Scotland no no yes yes no Constituency-specific time trends no no no no yes Adj. R-squared 0.58 0.50 0.52 0.74 0.83 Number of observations 1216 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. Standard errors are twoway-clustered on the constituency level and biannual level using the ivreg2 command in Stata. Discoveries (giant) denotes the number of giant oil fields discovered in t and t-1.

δ slightly decreases from 4.494 to 3.262, but remains significant at the 1-percent level. Column 3 relaxes the common trend assumption to some extent by adding a treatment-specific time trend which would capture any linear deviation from the common trend assumption. The coefficient decreases only slightly to 2.862 and remains significant at the 1-percent level. Column 4 adds the country-specific control variables and constituency-fixed effects. Note that any potential bias that could affect the treatment effect would have to occur at the country level, the constituency-fixed effects hence mainly serve to increase the efficiency of the estimations. Discoveries (giant) × Scotland still has a positive treatment effect of 1.923, significant at the 5- percent level. If our assumptions hold, inserting constituency-specific time trends should also not affect the coefficient, as these would only capture the treatment specific trend more precisely. Indeed, the coefficient remains almost identical: giant oil discoveries thus lead to an increase in the vote share of nationalist parties of about 2 percentage points.

This supports our hypothesis that nationalist parties can exploit the discoveries as a signal about the potential benefits of secession, which they try to communicate to voters and instrumentalize in their campaigns. It seems likely that voters also take the other years prior to the election into account when making their electoral choices. We examine this in two ways. First, we code variables that count the number of giant discoveries over the last 2, 3, 4, and 5 years prior to the election in year t. A priori, we would expect that 5 RESULTS 26 the effect is decreasing the longer the time period over which they accumulate. Voters most likely only incompletely remember all past events, so that a certain share will not take them into account in their optimization decision any more as time passes by. This is exactly what we observe: the treatment effect decreases from 1.923 in column 1 to 1.238 in column 5 (but the standard errors decrease even more). We interpret this as a sign that the salience of the discovery is important for its effect.

Table 3: Regression Results – Sum and Average Number of Oil Discoveries Dependent Variable: Nationalist vote share t={-1,0} t={-2,-1,0} t={-3,...,0} t={-4,...,0} t={-5,...,0} P Discoveries (giant)t × Scotland 1.923 1.555 1.365 1.309 1.238 [0.882] [0.483] [0.425] [0.390] [0.338] Discoveries per year (giant)t × Scotland 3.847 4.664 5.462 6.545 7.426 [1.765] [1.448] [1.701] [1.949] [2.028] t={-1,0} t={-2,-1,0} t={-3,...,0} t={-4,...,0} t={-5,...,0} P Discoveries (all)t × Scotland 0.718 0.585 0.521 0.483 0.399 [0.266] [0.193] [0.165] [0.135] [0.119] Discoveries per year (all)t × Scotland 1.437 1.754 2.083 2.414 2.396 [0.533] [0.580] [0.660] [0.676] [0.712] The table displays coefficients of 20 individual regressions with standard errors in brackets. All estimations also include constituency-fixed effects, biannual time-fixed effects, a linear time trend for Scotland as well as the control variables GDP per capita and Unemployment rate. Standard errors are twoway-clustered on the constituency level and biannual level using the ivreg2 command in Stata. I.e., the specifications are identical to column 4 in Table2. t= {−x, 0} denotes the sum/average number of (giant) oil discoveries in t and the x years prior to t. The number of observations is 1883.

While this prior test shows a decreasing coefficient for cumulative discoveries, we are interested in knowing whether voters react stronger if the number of additional discoveries per year is stable over the course of several years. A steady series of oil discoveries affirms voters that there are real potential benefits of secession. Row 2 in Table3 shows the coefficients of the individual regressions. The treatment effect increases from 3.487 for average discoveries in the election year and the year prior to election, to 7.426 if the number of discoveries per years is confirmed over the five years prior to an election. This is what we would expect: compared to considering the average discoveries over the last two years, an increase in one unit in average discoveries over the course of four years means that there were twice as many additional discoveries in total. It is also plausible that voters react more cautiously to single discoveries until further finds reduce the uncertainty about the long-term economic relevance of these regional resources on which the SNP campaigns. A graphical illustration of these results is presented in Figure 4.

Rows 3 and 4 in Table3 serve as a robustness tests using the count of all discoveries ( Discoveries (all) × Scotland in row 3, and Discoveries per year (all) × Scotland, row 4). While it is common in the literature to use giant oil discoveries which are more likely to be noticed (see, e.g., Lei & Michaels, 2014), it would increase our conﬁdence in the results if they hold for a larger sample of discoveries as well. We use all 5 RESULTS 27

Figure 4: Eﬀect of the Sum of Giant Discoveries and Giant Discoveries per Year

The upper graph shows the estimated coefficients and the respective 95-percent confidence interval from the first row in Table3. The lower graph displays the estimates from the second row reported in Table3. discoveries above 50 MMstb., as smaller ones are unlikely to have any economic significance. As expected, row 3 and 4 show lower coefficients compared to row 1 and 2. Anything else would have been surprising: as the average discovery is now much smaller in size, the additional effect of an additional discovery on Nationalist vote share should be smaller in a linear regression framework. However, all coefficients are significant at the 1-percent level. Moreover, they further support the pattern we observed before.

It seems intuitive that voters react to the number of discoveries, and more so to giant discoveries which cross a certain threshold to make it into the news regardless of being instrumentalized by the nationalist party or not. Still, it is also informative whether voters also react to the amount of oil that is discovered. 5 RESULTS 28

This is less obvious than it seems: while more oil is related to higher benefits of secession, it is unlikely that voters gather detailed information about the exact extent of the oil resources. The effect might thus not be linear in Amount of new reserves, or at least it is a less precise measure of the signal that the voters actually receive. Table4 shows the results focusing again first on the cumulative amount of oil discovered (row 1), i.e., in all discovered oil fields, and then the average amount of oil discovered per year. The coefficients remain positive throughout in both cases. They become statistically significant at the 1-percent level when the amount of oil discovered is confirmed over a course of at least three years. Again, the results lend support to our hypothesis that voters react more strongly when the signals are confirmed over a longer period of time. Exploring the coefficients and standard errors also indicates that the exact amount of oil discovered is a much noisier measure of what voters actually perceive. This is again in line with Lei & Michaels(2014) who argue that it is best to use the number of discoveries as a proxy. In terms of economic significance, discovering an additional 1000 MMstb. leads to an increase in the SNP’s vote share of about 2.5 percentage points.

Table 4: Regression Results – Amount of New Oil Reserves Dependent Variable: Nationalist vote share t={-1,0} t={-2,-1,0} t={-3,...,0} t={-4,...,0} t={-5,...,0} P Amount of new reservest × Scotland 0.735 0.787 0.655 0.532 0.499 [0.494] [0.277] [0.217] [0.168] [0.147] Amount of new reserves per yeart × Scotland 1.470 2.362 2.621 2.661 2.992 [0.988] [0.831] [0.869] [0.838] [0.882] The table displays coefficients of 10 individual regressions with standard errors in brackets. All estimations include constituency-fixed effects, biannual time-fixed effects, a linear time trend for Scotland as well as the control variables GDP per capita and Unemployment rate. Standard errors are twoway-clustered on the constituency level and biannual level using the ivreg2 command in Stata. I.e., the specifications are identical to column 4, Table2. t={−x, 0} denotes the sum/average amount of new discovered oil reserves in t and the x years prior to t. The number of observations is 1883.

To sum up the findings and their interpretations so far: we find a strong positive relationship between Discoveries (giant) and the vote share of the nationalist SNP party in a standard FE regression. The causal average treatment effect on the treated from the DiD design, using Wales as a control group, yields a slightly lower, but still highly significant coefficient. One additional giant oil discovery in the year of or the year prior to an election leads to an increase in Nationalist vote share of about 2 percentage points. We interpret this as the voters’ reactions to the change in the benefits of separation. Due to incomplete information, there is some uncertainty about the benefits, which are perceived as a noisy signal by the voters themselves or via party communication. Further tests taking into account several years prior to an election support the robustness of our results and show that the effect becomes stronger if there is a steady series of discoveries over a longer course of time. 5 RESULTS 29

5.1 Heterogeneous Eﬀects

This section examines whether there are heterogeneous effects across constituencies within Scotland. In Table5 we aim to assess which voters are more receptive to secessionist claims based on regional resources. We consider three areas on which heterogeneous effects could be based to learn more about voter behavior: differences in other benefits from a larger union, the salience of regional resources, and economic geography. Note that this goes beyond what our theoretical model captures, which assumes that factors like benefits from trade or preference heterogeneity are constant within regions.

If constituencies differ within regions, this could moderate the effect of increased regional resource value. Within a certain distance, commuting to work in England is feasible for people in Scotland or Wales and there will be more direct short-distance trade. Those voters are then involved in different discussions at work, get partly different media outlets, and are less exposed to SNP campaigns. To test this hypothesis, we coded whether a constituency was within 50, 75 or 100 km of the English border, and interacted these dummy variables with the treatment effect. Columns 1-3 show the results. The coefficient of the interaction term is about -1.5 for both the 50 and 75 km buffer, with standard errors around 0.5. In line with our expectations, the effect becomes smaller when our binary indicator also includes those within 75 to 100 km distance, decreases to -0.456 and becomes statistically indistinguishable from zero. Taking into account the main effect, the marginal effect is close to zero for those living in constituencies with close proximity to England. We take this as an indication for the importance of information acquisition and media exposure.

Secondly, we want to examine whether voters which are more directly affected by oil processing and whose jobs are potentially tied to the oil industry react differently with regards to nationalist party support. Column 4 shows the interaction of the treatment effect with a binary variable indicating coastal access of a constituency. This broadly captures whether there is a port from which ships could leave to the oil rigs and whether there are jobs related to the oil industry. While the main treatment effect remains positive, the positive interaction term with a point estimate of 1.716 and a standard error of 0.615 indicates that the effect of an oil discovery is higher if a constituency possesses coastal access. The simple binary measure still exhibits quite a bit of measurement error, e.g., as the Western coast of Scotland is not directly affected by oil regardless of its coastal access. Therefore, we also computed the distance to Aberdeen as a more precise measure and interact it in the same manner. Aberdeen is the main port serving offshore oil rigs and is often called the Oil Capital of Europe,39 with about half a million jobs being estimated to depend on the energy sector. The interaction term is negative with a point estimate of 0.009 and statistically highly significant.

39 See, e.g., BBC from 27th May 1997 at http://news.bbc.co.uk/2/hi/business/3236703.stm, last accessed on 31st Octo- ber 2016. 5 RESULTS 30 006 006] 339 846] 75 . . . . . 0 denotes [0 − 201291] - 508186] 2 [0 74 0 . . . . . 0 [0 − Discoveries (giant) 009004] - - 718076] 2 [1 75 0 . . . . . 0 [0 − 716615] - - - 831098] 5 [1 75 0 . . . . . [0 104 0 456686] - - -74 - 0 947] [1 . . . . . 0 [0 − 313 2 498423] - - - -75 - 0 848] [0 . . . . . 1 [0 − Regression Results 634417 2 577] - - - -75 - 0 - 870] [0 . . . . . 1 1 ------yesyesyes yes 1 yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [0 [0 − Table 5: vote share vote share vote share vote share vote share vote share vote share Nationalist Nationalist Nationalist Nationalist Nationalist Nationalist Nationalist . t-1 and t Ruggedness index Distance to Aberdeen Avg. soil suitability Near border (50) Near border (75) Near border (100) Coastal access × × × × × × × Discoveries (giant) Discoveries (giant) Discoveries (giant) Discoveries (giant) Discoveries (giant) Discoveries (giant) Discoveries (giant) Discoveries (giant) × × × × × × × × command in Stata. The main effects of the variables capturing potential heterogeneity are captured by the constituency-fixed effects. The table displays regression coefficients withivreg2 standard errors in brackets. Standardthe errors number are of twoway-clustered on giant the oil constituency fields level discovered and in biannual level using the Scotland Scotland Scotland Constituency-fixed effects Linear time trend Scotland Adj. R-squared 0 Scotland Scotland Scotland Scotland Scotland Biannual fixed effects Number of observations 1883 1883 1883 1883 1883 1883 1883 Dependent variable 5 RESULTS 31

Hence, the closer to Aberdeen a constituency is, the stronger the effect of additional oil discoveries. Note that this finding does not necessarily signal that people closer to Aberdeen care more about their pure self- interest. Whether their jobs would become more or less secure in an independent Scotland is uncertain. A more straightforward interpretation would be related to the issue salience hypothesis. According to that theory in political science, parties have issue reputations. I.e., in our context, when people think of North Sea Oil, the SNP is perceived as the party with the highest competence to handle this issue. The effect of an issue reputation on voter behavior is moderated by the attention and perceived importance of the respective issue (Bélanger & Meguid, 2008). Thus, one interpretation of our findings is that for voters with coastal access and those closer to Aberdeen the issue of oil revenues and their distribution is more salient. Hence, the positive effect of each discovery on secessionist party support is more pronounced. This indicates that the degree to which secessionist parties are able to instrumentalize regional resources might be moderated by the salience of the issue for voters.

Thirdly, we test whether the economic geography of constituencies affects our treatment effect. For that matter, we compute how suitable a district is to produce one of three main agricultural crops (potato, wheat, barley) and how rugged and therefore difficult to access and travel the constituency is. It seems possible that districts which are very suitable for agriculture would care less about revenues from other resources, and more rugged districts could be differentially affected for various reasons. In both cases, the interaction terms does not suggest that this influences the treatment effect. Both terms are negative, but far from conventional levels of significance. Thus, the effect of oil discoveries is not significantly altered by these considerations. Across the whole table, the main treatment effect always remains positive.

5.2 Addressing Potential Threats to Identiﬁcation

Finally, we want to consider remaining concerns regarding the causal interpretation of our results. While they are remarkably robust so far, some possible threats to identification exist. First, the distribution of the oil discoveries is skewed over time, with more discoveries in the 1970s than in later decades.40 Second, the relatively better development of the SNP since 1970 could coincide with a change to a better party leadership. Thirdly, while our main estimations cluster standard errors on the constituency and time level, other choices could also plausibly be argued for, which can be a critical choice in a DiD framework (Bertrand et al., 2004). Secessionist party success in each election can be correlated within a region, for instance if the party runs a successful campaign that might affect all constituencies. This is why we test for the robustness of the main results in Table 2 to alternative specifications. While an estimation with two clusters is not feasible, we can cluster on the region × time level. In addition, we categorize our

40 Note that our main results hold even when excluding the 1970s period completely, though. 5 RESULTS 32 sample period in five time categories and also cluster on the region × time-category level. This allows for arbitrary correlation of the error terms within the regions over approximately one decade. Both of these specifications lead to relatively few clusters, which is why we also simulate standard errors with a wild- cluster bootstrap procedure with 10,000 repetitions for these cluster levels in our two most conservative specifications (Cameron & Miller, 2015). Moreover, we cluster solely on the constituency or time level and use panel-corrected standard errors which model auto-correlation more specifically. In all cases, the treatment effects exhibits a p-value of 0.10 or less (see Appendix, Table 1-6).

To confirm that the effect we found is really driven by economic concerns , we make use of another plausibly exogenous variation. If the signal on which voters react is measuring the benefits of secession, as we hypothesized, the monetary value of the discovered oil should also be positively related to Nationalist vote share. We use the interaction between the average yearly world market price for oil and the amount of discovered oil times the Scotland dummy to test this hypothesis.41 While there could be a small correlation between Scottish oil discoveries and the oil price, we follow the bulk of the existing literature and treat changes in the world market oil price as exogenous (Arezki & Brückner, 2012, 2011). One potential concern would be that both supply and demand in Scotland are related to both the oil price and secessionist party success. While this is a valid concern for large producers like the Arab countries, or countries that represent a sizeable share of world demand like the US (Kilian & Park, 2009) it seems justifiable that the effect of variations in Scottish oil production and demand are negligible over the course of a year. If regional resource value is driving nationalist party success, oil discoveries should matter more when the oil price is higher, i.e., when their net value as a potential benefit of secession is larger.

The regression equation now becomes:

× × Nationalist Vote Sharei,c,t = α + δ Discoveries (giant)t Scotlandc,i + θ Scotlandc,i Pricet + η × × + 0 γ + λ + ϑ + τ + ε Discoveries (giant)t Scotlandc,i Pricet Xi,t i t cT c,i,t , and we focus on the triple interaction coeﬃcient η.

Table6 shows the results of these triple-differences specifications. Row 1 uses the amount of discoveries, and row 2 the amount of discovered oil. Columns 1-5 show consistently positive interaction effects for different intervals, which are statistically significant at least at the 5-percent level. The same holds for the interaction with the amount of discovered oil. The positive interaction effects are all statistically significant at least at the 1-percent level. Voters react more strongly when the “price” they receive for secession at the moment

41 We choose the price for Brent Crude which is suitable for North Sea oil. Brent trades at a higher price than the other major classiﬁcation West Texas Intermediate (WTI), but as both are close substitutes, the prices are strongly correlated. Even if one assumes that the oil price is endogenous, we can interpret the interaction between an exogenous variable and an endogenous variable as causal under relatively mild assumptions (c.f., Dreher et al., 2016). 5 RESULTS 33

Table 6: Regression Results – Oil Price Dependent Variable: Nationalist vote share t={-1,0} t={-2,-1,0} t={-3,...,0} t={-4,...,0} t={-5,...,0} P Discoveries (all)t × Scotland × Oil price 0.014 0.013 0.012 0.016 0.009 [0.006] [0.005] [0.004] [0.005] [0.004] P Amount of new reservest × Scotland × Oil price 0.021 0.021 0.016 0.013 0.012 [0.009] [0.008] [0.006] [0.004] [0.004] The table displays coefficients of 10 individual regressions with standard errors in brackets. All estimations include constituency- fixed effects, biannual time-fixed effects, a linear time trend for Scotland as well as the control variables GDP per capita and Unemployment rate. The main effect of Scotland × Oil price is also included, but not displayed. Standard errors are twoway- clustered on the constituency level and biannual level using the ivreg2 command in Stata. I.e., the specifications are identical to column 4, Table2. t={−x, 0} denotes the sum/average amount of new discovered oil reserves in t and the x years prior to t. The number of observations is 1883. of voting is more lucrative. This is in line with the literature on the nexus between development aid or natural resources and conflict, where an increased “price” is linked to more separation conflicts (Morelli & Rohner, 2015). Obviously, very similar mechanisms are at play within a democratic system with open elections. Translated into vote shares, if the oil price is $20 higher, each additional discovery is worth 0.38 percentage points more, or alternatively an additional MMstb. is worth 0.42 percentage points more for the SNP.42

The triple-differences design offers another advantage. We can make use of it to implement a placebo test which also implicitly tests the DiD assumptions: if the differences between Scotland and Wales are really caused by an increase in the value of regional resources and not by some unobserved other factor, we would expect that the oil price has a positive effect after the first discovery, but to observe no effect before the first oil was discovered. Table7 hence shows the results of two models: the first column includes all observations before the discovery of the first oil field in 1970; the second column covers all observations from 1970 onwards.

The results are very reassuring: before 1970, there is no significant positive effect of the oil price for Scot- land. In fact the coefficient is negative, but far from any conventional level of significance. At the same time, the coefficient is positive and highly significant in the period after the first discovery. This further confirms our confidence in the causal interpretation of our results. The SNP succeeds in instrumentaliz- ing the oil discoveries to gain about 2 percentage points more votes per giant oil discovery. Accordingly,

42 Note that, in accordance with our prior approach, we use new discoveries instead of the stock of discovered oil. The reason is mostly to remain consistent with the earlier results. In addition, we find it more plausible that voters react to changes in the value of newly discovered oil. Remembering or estimating the cumulative amount of discovered oil and subtracting already exploited oil imposes higher search and information costs on the voters. In line with our results on heterogeneity between constituencies, we could also say that new discoveries are more salient due to their media coverage. Still, we also compute the amount of discovered and unexploited oil for each point in time. The interaction with the indicator for Scotland and the oil price is positive in the whole sample and becomes highly significant when only using the time period after the first oil discovery when the price should matter. 6 CONCLUSIONS 34

Table 7: Pre- and Post-1970 Effect of the Oil Price Dependent Variable: Nationalist vote share Nationalist vote share Observation period: 1945-1969 1970-2001 Oil price × Scotland −0.268 0.125 [0.251] [0.045] p-value: Oil price × Scotland 0.286 0.005 Adj. R-squared 0.54 0.81 Number of observations 841 1042 The table displays regression coefficients with standard errors in brackets. The first (giant) oil discoveries were in 1970. Both estimations include constituency-fixed effects, biannual time-fixed effects, as well as the control variables GDP per capita and Unemployment rate. Standard errors are twoway-clustered on the constituency level and biannual level using the ivreg2 command in Stata. resource wealth can not only lead to secessionist conflicts in developing countries (see, e.g., Collier, 2010; Ross, 2004a), they also contribute to the success of secessionist parties in established democracies.

6 Conclusions

Our paper augments the existing literature on the size of nations and sheds light on the factors that determine the success of secessionist parties. The main argument is that citizens take the value of regional resources into account when deciding whether to support secession or not. Secessionist parties can suc- cessfully exploit regional resources to increase their vote share. Based on a simple theoretical model, we demonstrate with numerous cases that there clearly exists such a positive correlation in various countries and regions. Importantly, we employ a broader deﬁnition of resources, which does not only comprise natural resources, but also the value of geographically concentrated human capital, land suitability and other factor endowments. Building on these cases, we turned to the United Kingdom to test whether we can establish a causal relationship between resource value and secessionist party vote share.

As we argue, Scotland and Wales are suitable counterfactuals, so that we can use the discoveries of North Sea oil as a natural experiment. Our constituency level analysis covering all UK elections over the 1945-2001 period shows that Scottish voters react in a significant way to oil discoveries which increase the perceived benefits of secession. In a DiD setting with Plaid Cymru in Wales as the counterfactual, the vote share of the Scottish National Party, the main advocate of a secession of Scotland from the UK, significantly increased by about 2 percentage points after the discovery of giant oil fields off the Scottish coast. We argue in-depth why the Welsh Plaid Cymru and the Scottish SNP form a suitable control and treatment group with no visible differences in pre-trends. Besides this quantitative argumentation, we also provided qualitative arguments 6 CONCLUSIONS 35 why both parties are comparable and followed a similar trend prior to the first oil discovery. Historical evidence clearly indicates that at the time of the first discovery, it was a mostly unexpected event.

Overall, this adds an important dimension to the literature on democratic secessionism and the size of nations. It also connects the literature from the developing country context with established democracies. Showing a causal effect of a change in regional resource wealth indicates that it is not solely ethnic or cultural differences that determine the strength of democratic secessionism. Based on the evidence from Sorens(2005) and Van Houten(2007), cultural factors like language and the weak dynamics of ethnic group affiliation may define a certain baseline but do not seem to account for the larger part of the variation in support for Scottish independence. Brand et al. (1994, p. 629) put this in a nutshell: “If the SNP were to emphasize its Scottishness over its concern for the prosperity of the country, it would lose the vast majority of its voters, members, and probably most of its leaders.” The numerous other examples where changes in regional resource value correlate with secessionist party success indicate that the mechanisms we outlined matter beyond this one case which allowed as a clear causal identification.

Certain requirements for this to matter have to be met. First, the resource value must be so signiﬁcant that it alters the costs and beneﬁts of secession in a sizeable way. Second, it must be geographically concentrated in a region that regards itself as a group with some kind of pre-existing common group identity on which a campaign can be built. Third, the economic gains from the respective regional resource are to some degree being re-distributed within the country. Exploring these questions in more detail, and also evaluating on the choice between secession or decentralization, should provide a fruitful area for future research.

Within Scotland, the failed referendum in 2015 indicates that in times of low oil prices the mere cultural diﬀerences are not enough to convince voters of the beneﬁts of secession. The future will show whether a potential second attempt will prove more successful for the Scots. Ironically, there might be another dynamic developing in such a case. As some Westminster politicians like to argue, the Shetland Islands could subsequently aim at a secession from mainland Scotland motivated by the fact that large parts of the oil resources actually lie within their theoretical maritime borders. Hence, regional resources and their distribution will continue to matter in the near and distant future. In light of these potentially turbulent future secession dynamics, we revisit and continue the initial quotation from the The Parable of the Wise and Foolish Virgins:

“Therefore keep watch, because you do not know the day or the hour.” – Matthew 25, 13 REFERENCES 36

References

Alesina, Alberto, & Spolaore, Enrico. 1997. On the Number and Size of Nations. The Quarterly Journal of Economics, 112(4), 1027–1056.1,5,6

Alesina, Alberto, Perotti, Roberto, & Spolaore, Enrico. 1995. Togetheror Separately? Issues on the Costs and Beneﬁts of Political and Fiscal Unions. European Economic Review, 39(3), 751–758.6

Arezki, Rabah, & Brückner, Markus. 2011. Oil Rents, Corruption, and State Stability: Evidence from Panel Data Regressions. European Economic Review, 55(7), 955–963.1, 32

Arezki, Rabah, & Brückner, Markus. 2012. Commodity Windfalls, Democracy and External Debt. The Economic Journal, 122(561), 848–866.1, 32

Arezki, Rabah, Ramey, Valerie A., & Sheng, Liugang. 2015. News Shocks in Open Economies: Evidence from Giant Oil Discoveries. NBER Working Paper No. 20857. 18, 21

Ashbrook, John, & Ashbrook, John E. 2008. Buying and Selling the Istrian Goat: Istrian Regionalism, Croatian Nationalism, and EU Enlargement. Peter Lang. 10

Baum, Christopher F., Schaﬀer, Mark E., & Stillman, Steven. 2010. Ivreg2: Stata Module for Extended Instru- mental Variables/2SLS and GMM Estimation. http://ideas.repec.org/c/boc/bocode/s425401. html. 24

Baumeister, Christiane, & Peersman, Gert. 2013. The Role of Time-Varying Price Elasticities in Accounting for Volatility Changes in the Crude Oil Market. Journal of Applied Econometrics, 28(7), 1087–1109. 19

Bélanger, Éric, & Meguid, Bonnie M. 2008. Issue Salience, Issue Ownership, and Issue-based Vote Choice. Electoral Studies, 27(3), 477–491. 31

Bertrand, Marianne, Duflo, Esther, & Mullainathan, Sendhil. 2004. How Much Should We Trust Differences-in-Differences Estimates? Quarterly Journal of Economics, 119(1). 31

Boﬀa, Federico, Piolatto, Amedeo, & Ponzetto, Giacomo A. M. 2016. Political Centralization and Gov- ernment Accountability. The Quarterly Journal of Economics, 131(1), 381–422.5

Bolton, Patrick, & Roland, Gerard. 1997. The Breakup of Nations: A Political Economy Analysis. The Quarterly Journal of Economics, 112(4), 1057–1090.5,6

Bordignon, Massimo, & Brusco, Sandro. 2001. Optimal Secession Rules. European Economic Review, 45(10), 1811.5 REFERENCES 37

Brancati, Dawn M. 2015. Global Elections Database. http://www.globalelectionsdatabase.com. 17

Brand, Jack, Mitchell, James, & Surridge, Paula. 1994. Social Constituency and Ideological Proﬁle: Scottish Nationalism in the 1990s. Political Studies, 42(4), 616. 17, 35

Brender, Adi, & Drazen, Allan. 2008. How Do Budget Deﬁcits and Economic Growth Aﬀect Reelection Prospects? Evidence from a Large Panel of Countries. The American Economic Review, 98(5), 2203–2220. 5

Brocklehurst, Steven. 2013. Who Has a Right to Claim North Sea Oil? BBC Scotland News, http://www. bbc.co.uk/news/uk-scotland-scotland-politics-20042070. 13, 15

Brown, J. Gordon. 1975. The Red Paper on Scotland. Edinburgh: EUSPB. 15

Bryant, Christopher G. A. 2006. The Nations of Britain. 1. publ. edn. Oxford: Oxford Univ. Press. 13, 19, 22

Buchanan, James M., & Faith, Roger L. 1987. Secession and the Limits of Taxation: Toward a Theory of Internal Exit. The American Economic Review, 77(5), 1023.5

Cameron, A. Colin, & Miller, Douglas L. 2015. A Practitioners Guide to Cluster-Robust Inference. Journal of Human Resources, 50(2), 317–372. 32

Caselli, Francesco, Morelli, Massimo, & Rohner, Dominic. 2015. The Geography of Interstate Resource Wars. The Quarterly Journal of Economics, 130(1), 267–315.3

Casertano, Stefano. 2013. Natural Resources and the Question of Ethnic Nationalisms. Pages 253–267 of: Our Land, Our Oil! Wiesbaden: Springer.3,4,9

Cento Bull, Anna, & Gilbert, Mark. 2001. The Lega Nord and the Northern Question in Italian Politics. Palgrave. 10

Collier, Paul. 2010. The Political Economy of Natural Resources. Social Research, 77(4), 1105–1132.1, 34

Collier, Paul, & Hoeﬄer, Anke. 2004. Greed and Grievance in Civil War. Oxford Economic Papers, 56(4), 563–595.3

Collier, Paul, & Hoeﬄer, Anke. 2006. The Political Economy of Secession. Pages 37–59 of: Hannum, Hurst, & Babbitt, Eileen (eds), Negotiating Self-Determination. Lanham, Maryland: Lexington Books.2, 3,4,9, 16, 20, 21, 23

Dehdari, Sirus, & Gehring, Kai. 2016. Building National Identity? The Causal Eﬀect of Homogenization Policies in Alsace-Lorraine. CIS Working Paper No. 89.8 REFERENCES 38

Dixit, Avinash, & Londregan, John. 1998. Fiscal Federalism and Redistributive Politics. Journal of Public Economics, 68(2), 153–180.1,7

Dreher, Axel, Eichenauer, Vera Z., & Gehring, Kai. 2016. Geopolitics, Aid, and Growth: the Impact of UN Security Council Membership on the Eﬀectiveness of Aid. The World Bank Economic Review. 32

Eerola, Essi, Määttänen, Niku, & Poutvaara, Panu. 2004. Citizens should Vote on Secession. Topics in Economic Analysis & Policy, 4(1).8

Fusaro, Anthony. 1979. Two Faces of British Nationalism: The Scottish National Party & Plaid Cymru Compared. Polity, 11(3), 362–386. 22, 23

Gehring, Kai, & Schneider, Stephan A. 2016. Towards the Greater Good? EU Commissioners’ Nationality and Budget Allocation in the European Union. CIS Working Paper No. 86.6

Gourevitch, Peter A. 1979. The Reemergence of “Peripheral Nationalisms”: Some Comparative Specula- tions on the Spatial Distribution of Political Leadership and Economic Growth. Comparative Studies in Society and History, 21(3), 303–322. 13

Goyal, Sanjeev, & Staal, Klaas. 2004. The Political Economy of Regionalism. European Economic Review, 48(3), 563.5

Harvie, Christopher. 1995. Scotland and Nationalism. 2. ed., repr. edn. London: Routledge.2, 17

Heald, David. 1994. Territorial Public Expenditure in the United Kingdom. Public Administration London, 72, 147–147. 19

Hechter, Michael. 1992. The Dynamics of Secession. Acta Sociologica, 35(4), 267–283.9

Holitscher, Marc, & Suter, Roy. 1999. The Paradox of Economic Globalisation and Political Fragmentation: Secessionist Movements in Quebec and Scotland. Global Society: Journal of Interdisciplinary International Relations, 13(3), 257–286.2, 13

Horowitz, Leah S. 2004. Toward a Viable Independence? The Koniambo Project and the Political Economy of Mining in New Caledonia. The Contemporary Paciﬁc, 16(2), 287–319. 11

Ivanhoe, L. F. Buz, & Leckie, George G. 1993. Global Oil, Gas Fields, Sizes Tallied, Analyzed. Oil & Gas Journal, 91(7), 87–91. 18

Jolly, Seth K. 2007. The Europhile Fringe? European Union Politics, 8(1), 109–130. 16, 21

Jolly, Seth K. 2014. Voting for Nation or State: Determinants of Independence Support in Scotland & Catalonia. APSA 2014 Annual Meeting Paper.8 REFERENCES 39

Jones, Richard W., & Scully, Roger. 2006. Devolution and Electoral Politics in Scotland and Wales. The Journal of Federalism, 36(1), 115–134. 17

Jordahl, Henrik. 2006. An Economic Analysis of Voting in Sweden. Public Choice, 127(3/4), 251–273.5

Kemp, Alexander G., & Stephen, Linda. 2000. Expenditures in and Revenues from the UKCS: Estimating the Hypothetical Scottish Shares. North Sea Study Occasional Paper No. 70. 13, 15, 21

Kemp, Alexander G., Hallwood, C. Paul, & Wood, Peter W. 1983. The Beneﬁts of North Sea Oil. Energy Policy, 11(2), 119–130. 15

Kilian, Lutz, & Park, Cheolbeom. 2009. The Impact of Oil Price Shocks on the US Stock Market. Inter- national Economic Review, 50(4), 1267–1287. 32

Kuklinski, James H., & West, Darrell M. 1981. Economic Expectations and Voting Behavior in United States House and Senate Elections. American Political Science Review, 75(2), 436–447.5

Lee, Clive H. 1995. Scotland and the United Kingdom: The Economy and the Union in the Twentieth Century. Manchester: Manchester Univ. Press. 15, 19, 22

Lee, Geoﬀrey W. 1976. North Sea Oil and Scottish Nationalism. Political Quarterly, 47(3), 307–317. 15, 23

Lei, Yu-Hsiang, & Michaels, Guy. 2014. Do Giant Oilﬁeld Discoveries Fuel Internal Armed Conﬂicts? Journal of Development Economics, 110, 139–157.3, 18, 21, 26, 28

Levy, Roger. 1995. Finding a Place in the World-Economy. Party Strategy and Party Vote: The Regional- ization of SNP and Plaid Cymru Support, 1979-92. Political Geography, 14(3), 295.2, 22, 23

Lewis, M. Paul, Simons, Gary F., & Fennig, Charles D. 2009. Ethnologue: Languages of the Eorld. Vol. 16. SIL international Dallas, TX. 12

Lewis-Beck, Michael S., & Stegmaier, Mary. 2000. Economic Determinants of Electoral Outcomes. Annual Review of Political Science, 3, 183–219.4,5

Lynch, Peter. 2003. Marriage versus Divorce: Political Parties and the Costs and Beneﬁts of Secession and Union in Scotland. Canadian Political Science Association Annual Meeting, Dalhousie University, Halifax.2, 7, 15, 16, 19

MacKay, Donald I., & Mackay, George A. 1975. The Political Economy of North Sea Oil. London: Robert- son.2, 14, 15, 16, 21

McCrone, David. 1992. Understanding Scotland. 1. publ. edn. International Library of Sociology. London: Routledge. 13 REFERENCES 40

McGuinness, Feargal, Cracknell, Richard, & Taylor, Mark. 2012. UK Election Statistics: 1918-2012. Com- mons Brieﬁng Papers No. RP 12-43.2, 16, 17

Miller, William L. 1981. The End of British Politics? Oxford: Clarendon Press. 16

Miller, William L., Sarlvik, Bo, Crewe, Ivor, & Alt, Jim. 1977. The Connection Between SNP Voting and the Demand for Scottish Self-Government. European Journal of Political Research, 5(1), 83–102. 16, 17

Mitchell, Brian R. 1988. British Historical Statistics. 1. publ. edn. Cambridge: Cambridge Univ. Press. 19

Mitchell, James, Bennie, Lynn, & Johns, Rob. 2012. The Scottish National Party: Transition to Power. Oxford: Oxford Univ. Press. 16, 17, 22

Mnookin, Robert, & Verbeke, Alain. 2009. Persistent Nonviolent Conﬂict with no Reconciliation: The Flemish and Walloons in Belgium. Law and Contemporary Problems, 72(2), 151–186. 12

Morelli, Massimo, & Rohner, Dominic. 2015. Resource Concentration and Civil Wars. Journal of Develop- ment Economics.3,4, 33

Nagel, Klaus-Jürgen. 2004. Transcending the National / Asserting the National: How Stateless Nations like Scotland, Wales and Catalonia React to European Integration. Australian Journal of Politics & History, 50(1), 57–74.6, 23

Outlaw, Ian. 2012. UK General Election Results 1992. http://www.politicsresources.net/area/uk/ ge92/ge92index.htm. 17

Paquin, Stéphane. 2002. Globalization, European Integration and the Rise of Neo-Nationalism in Scotland. Nationalism and Ethnic Politics, 8(1), 55–80.6, 13, 16

Persson, Torsten, & Tabellini, Guido. 1996. Federal Fiscal Constitutions: Risk Sharing and Redistribution. Journal of Political Economy, 979–1009.1,7

Persson, Torsten, & Tabellini, Guido Enrico. 2002. Political Economics: Explaining Economic Policy. MIT Press.1,7

Ploeg, Frederick. 2011. Natural Resources. Journal of Economic Literature, 49(2), 366–420.3

Popiden, Sandra. 2011. Globalization, Natural Resources, and Separatist Wars. http://papers.ssrn.com/ sol3/papers.cfm?abstract_id=1919919.4

Powell, G. Bingham, & Whitten, Guy D. 1993. A Cross-National Analysis of Economic Voting: Taking Account of the Political Context. American Journal of Political Science, 37(2), 391–414.4 REFERENCES 41

Ross, Michael L. 2004a. How Do Natural Resources Inﬂuence Civil War? Evidence from Thirteen Cases. International Organization, 58(1), 35–67.1,4, 34

Ross, Michael L. 2004b. What Do We Know about Natural Resources and Civil War? Journal of Peace Research, 41(3), 337–356.3

Rosser, Andrew. 2006. The Political Economy of the Resource Curse: A Literature Survey. IDS Working Paper No. 268.1,3

Scott, Peter. 2004. Regional Development and Policy. Pages 332–367 of: Floud, Roderick, & Johnson, Paul (eds), The Cambridge Economic History of Modern Britain – Structural Change and Growth, 1939 - 2000, 1. publ. edn., vol. 3. Cambridge: Cambridge Univ. Press. 15, 19

Sorens, Jason. 2004. Globalization, Secessionism, and Autonomy. Electoral Studies, 23(4), 727–752.7, 16

Sorens, Jason. 2005. The Cross-Sectional Determinants of Secessionism in Advanced Democracies. Com- parative Political Studies, 38(3), 304.5,8,9, 16, 17, 19, 22, 35

Tufte, Edward R. 1978. Political Control of the Economy. 2. print. edn. Princeton, NJ: Princeton Univ. Press.4

UK Statistics Authority. 2013. Oﬃcial Statistics in the Context of the Referendum on Scottish Indepen- dence. Monitoring Report No. 6. 14, 15, 21

Van Houten, Pieter. 2007. Regionalist Challenges to European States: A Quantitative Assessment. Ethnop- olitics, 6(4), 545–568.7, 35

Whaley, Jane. 2010. The First UK Giant Oil Field. GeoExPro, 7(3), 76–80. 14, 20

Wood, John R. 1981. Secession: A Comparative Analytical Framework. Canadian Journal of Political Science, 14(1), 107–134.5,9

Wooldridge, Jeﬀrey M. 2010. Econometric Analysis of Cross Section and Panel Data. 2. ed. edn. Cambridge, Mass.: MIT Press. 21, 24

Young, Robert. 2014. Transition Costs in Secessions, with a Brief Application to Scotland. Oxford Review of Economic Policy, 30(2), 392–405.6,7

Zirakzadeh, Cyrus Ernesto. 1989. Economic Changes and Surges in Micro-Nationalist Voting in Scotland and the Basque Region of Spain. Comparative Studies in Society and History, 31(2), 318–339. 19 42

Appendix

Contents

Appendix 42

A Region Proﬁles...... 43

B Giant Discoveries...... 64

C Calculation of Constituency Results...... 67

D Variables Description...... 76

E Estimated Scottish Shares...... 78

F Regressions...... 79

G It’s Scotland’s Oil!...... 85 A Region Proﬁles 43

A Region Proﬁles

Flanders

• Seeking independence from Belgium

• Political parties: New Flemish Alliance (secessionist), Libertair Direct Democratisch (secessionist), Vlaams Belang (secessionist), Identity, Tradition, Sovereignty (secessionist, defunct since 2007)

Logo of the New Flemish Alliance (Nieuw-Vlaamse Alliantie)

• Resources: Flanders was the poorer region in Belgium up until the Second World War, as it was the last Belgian region to industrialize, and relied strongly on the agricultural sector (1). This changed radically in the past 60 years due to Flanders’ successful transformation to a knowledge-based economy with highly developed service and high-tech sectors (2). Today, supposedly 6 billion EUR per year are transferred to Wallonia and Brussels (3). Antwerp is home to the second largest European sea port by cargo volume and per capita GDP is 32,700 EUR compared to Wallonia’s 26,100 EUR (2).

Electoral success and party strategies: The secessionist New Flemish Alliance presents the high regional transfers to Wallonia and Brussels as a key argument for independence, devoting an entire brochure titled “Vlaanderen betaalt de Belgische factuur” (“Flanders pays the Belgian bill”) to the topic. More- over, the support for regionalist and separatist parties steadily increased from only 5 percent in 1961 to almost 45 percent in 2010, correlating with Flanders’ economic rise and a steady increase in regional transfers, despite the granting of equal linguistic rights in the 1950s (4). In particular, the abrupt increase from 7.3 to 7.9 billion EUR following the 2008 financial crisis goes hand in hand with a sharp upsurge in support for regionalist and secessionist parties (see figures below). A Region Profiles 44

Source: Deschouwer(2013, p, 349)

Source: “Flanders Pays the Belgian Bill”, New Flemish Alliance at https://www.n-va.be/sites/default /files/generated/files/brochure-attachment/brochure_vlaanderen_betaalt_de_belgische_f actuur.pdf (p. 12), last accessed on 31st October 2016

• Sources: (1) See the Financial Times from 3rd November 2015 at https://www.ft.com/content/c45dfb d4-7349-11e5-bdb1-e6e4767162cc, last accessed on 31st October 2016 (2) See Knowledge@Wharton from 2nd December 2013 at http://knowledge.wharton.upenn.ed u/article/secession-answer-case-catalonia-flanders-scotland/, last accessed on 31st Oc- tober 2016 (3) See Hermans(2015) A Region Proﬁles 45

• Quotes: “The most dramatic example [of economic contrast] is in Belgium, where the growing gap between Flanders and French-speaking Wallonia has exacerbated political and cultural tensions. The NVA party, which rules Flanders, believes that wealthy Flanders should not be subsidizing poorer Wallonia, whose regional government is alleged to be wasting money. Flemish nationalists feel strongly that their region is not receiving its fair share of the revenues that it contributes to the national economy.” (Knowledge@Wharton from 2nd December 2013 at http://knowledge.wharton.upenn.edu/ar ticle/secession-answer-case-catalonia-flanders-scotland/, last accessed on 31st Octo- ber 2016)

“To this strong Flemish identity, an economic component has also been added over the course of recent decades. During the nineteenth and the ﬁrst half of the twentieth century Wallonia was the economically stronger region. That changed after the Second World War as a result of industrial decline in the south and the development of new economic activities in the north. Today, Flanders is the stronger region. However, the relative wealth of Flanders, combined with the operation of the welfare state put into place after the Second World War, has meant that a system of social redistribution has eﬀectively become a system of territorial redistribution. When one aggregates per region the amount of money paid into the system and the amount of money received from the system, Flanders is a net contributor and Wallonia (and increasingly also Brussels) is a net recipient” (Deschouwer, 2013).

“Billions of euros in transfers are going to from Flanders to Wallonia and Brussels. And yet or- dinary people in Wallonia and Brussels are not better oﬀ because of them. And the worse it gets for them, the higher the transfers are. Policymakers are therefore not at all encouraged to even change their actions. Achieving improvement inevitably means: less transfers, less money.” (“Flanders Pays the Belgian Bill”, New Flemish Alliance at https://www.n-va.be/sites/default /files/generated/files/brochure-attachment/brochure_vlaanderen_betaalt_de_belgi sche_factuur.pdf, last accessed on 31st October 2016) A Region Proﬁles 46

Wallonia

• Seeking independence from Belgium

• Political parties: Rassemblement Wallonie France (formerly federalist, secessionist since 1985)

Logo of the Rassemblement Wallonie France (Rally Wallonia France)

• Resources: Wallonia was the the ﬁrst Belgian region to industrialize in the 19th century (1), rendering it the richer part of Belgium up to the 1960s due to comparative advantages in steel production and coal mining (2). The steel crises of the 1970s and the general decline of the heavy industries in Europe caused Wallonia to experience strong economic decline(2). Today, Wallonia is the signiﬁcantly poorer region with a per capita GDP of only 26,100 EUR compared to Flanders’ 32,700 EUR and receives high transfers from Flanders (3).

Electoral success and party strategies: The first Walloon independence movements emerged in the 19th century, coinciding with the region’s industrialization. However, unlike the Flemish nationalist parties, pro-independence parties in Wallonia never gained significant traction. One possible explanation in line with our theory is that although Wallonia was the significantly richer region up to the 1960s, there never was a perceived economic benefit of secession for Walloons. This is due to the fact that “[n]either in the nineteenth, nor in the twentieth century did a Walloon tax surplus flow to Flanders", as the Flemish historian Prof. em. Juul Hannes postulates (4), which can be explained by the absence of a welfare state prior to the Second World War. The construction of the welfare state in the post-war area in effect imposed a system of regional redistribution, with a Flemish tax surplus of approximately 150 million EUR flowing to Wallonia as early as 1955 (4). Accordingly, the Rassemblement wallon (RW), the main pro-autonomy party in the 20th century, received only 7 percent of the vote at its peek in the 1970s. Today, the Walloon Rally usually stays below the 2 percent mark in federal elections (5).

• Sources: (1) See the Financial Times from 3rd November 2015 at https://www.ft.com/content/c45dfb A Region Proﬁles 47

d4-7349-11e5-bdb1-e6e4767162cc, last accessed on 31st October 2016 (2) See, e.g., Reid & Musyck(2000) and Witte(1992) (3) See Knowledge@Wharton from 2nd December 2013 at http://knowledge.wharton.upen n.edu/article/secession-answer-case-catalonia-flanders-scotland/, last accessed on 31st October 2016 (4) “Flanders Pays the Belgian Bill”, New Flemish Alliance at https://www.n-va.be/sites/defau lt/files/generated/files/brochure-attachment/brochure_vlaanderen_betaalt_de_bel gische_factuur.pdf, last accessed on 31st October 2016 (5) Duerr(2016, p. 12)

• Quotes: “In the century and a half up to the 1960s, the Walloon economy was one of the most prosperous in Europe. [...] At the time of the ﬁrst industrial revolution, Wallonia was equipped with numerous comparative advantages in the leading sectors of the epoch: coal mining, steel making and their spin-oﬀ activities. Natural resources, a highly skilled workforce and the dynamism of its engineers were the foundations on which Wallonia built its prosperity.” (Reid & Musyck, 2000, p. 183)

“Wallonia was among the ﬁrst regions in northern Europe to industrialise in the 19th century, with industries such as glass making and coal mining. By contrast, the largely agrarian Flanders fell behind. But Flanders boomed in the postwar era, attracting much foreign investment. The Walloon economy, meanwhile, collapsed as the region’s main heavy industries faltered. Between 1980 and 2010, the number of jobs in manufacturing halved from one in four to just one in 10.” (Financial Times from 3rd November 2015 at https://www.ft.com/content/c45dfbd4-7349-11e 5-bdb1-e6e4767162cc, last accessed on 31st October 2016)

“The region’s economic numbers are dire. Wallonia’s share of GDP is small and heading in the wrong direction. The region counts for a third of Belgium’s 11m population but less than a quarter of its GDP – and this number is falling.” (Financial Tomes from 6th November 2014 https://www.ft.com/content/7ee4c346-52e1-11e 4-9221-00144feab7de, last accessed on 31st October 2016)

“The unequal economic situation is one of the most striking aspects of this. Wallonia still has to face up to the problems of restructuring its old branches of industry and the Walloon economy has done relatively little towards setting up ’high-tech’ sectors”(Witte, 1992, p. 109). A Region Proﬁles 48

Catalonia

• Seeking independence from Spain

• Political parties: Republican Left of Catalonia (secessionist), Democratic Convergence of Catalonia (secessionist), Popular Unity Candidacy (secessionist)

Logo of the Esquerra Republicana de Catalunya (Republican Left of Catalonia)

• Resources: Historically, Catalonia was among the ﬁrst Spanish regions to industrialize and featured a strong industrial as early as the beginning of the 19th century. Today, the region sets itself apart from the rest of Spain as the richest and most successful exporting region. Exports generate 28.1 percent of the regional GDP, compared with just 12 percent in Madrid. A new record was reached in 2012, with exports amounting to 58.2 billion EUR which is 15.4 percent higher than before the economic crisis (1).

Electoral success and party strategies: The economic crisis has strengthened resentment towards the Spanish system of regional redistribution which annually transfers 8 percent to 9 percent of Catalo- nia’s GDP to less prosperous Spanish regions (1, 3). Secessionist parties like Esquerra (Republican Left of Catalonia) argue that Catalonia would beneﬁt from complete ﬁscal autonomy, as part of Catalonia’s debt can be blamed on the “wasteful central state” (4).

• Sources: (1) See Knowledge@Wharton from 2nd December 2013 at http://knowledge.wharton.upen n.edu/article/secession-answer-case-catalonia-flanders-scotland/, last accessed on 31st October 2016 (2) See Instituto Nacional de Estadística at http://www.ine.es/en/daco/daco42/cre00/b2010/ homog/dacocre_base2010h_en.htm, last accessed on 31st October 2016 (3) See the Financial Times from 26th September 2012 at https://www.ft.com/content/bad 90798-07f4-11e2-9df2-00144feabdc0, last accessed on 31st October 2016 (4) See Election Manifesto 2016, Republican Left of Catalonia (p. 4) at http://www.esquerra.cat A Region Proﬁles 49

/partit/programes/e2016-programa.pdf, last accessed on 31st October 2016

• Quotes: “We suffer from the effects of a wasteful central state that, in addition to a 16,000 million annual fiscal deficit, throws out our resources for the AVE [high-speed rail in Spain] without passengers, airports without airplanes and military spending. We want a welfare state for ourselves, managing our resources and to ensure the construction of the infrastructure, because we need to go forward.” (Election Manifesto 2016, Republican Left of Catalonia (p. 4) at http://www.esquerra.cat/parti t/programes/e2016-programa.pdf, last accessed on 31st October 2016)

“A majority of Catalans feels Madrid takes too much of local income to redistribute elsewhere. The clamour for independence has become mainstream.” (Financial Times from 26th September 2012 at https://www.ft.com/content/bad90798-07f 4-11e2-9df2-00144feabdc0, last accessed on 31st October 2016)

“The perception that an independent Catalonia would perform better economically, based on the idea that the current ﬁscal relationship is detrimental to Catalonia’s interests, partly explains current support for independence.” (Muñoz & Tormos, 2012, p. 316) A Region Proﬁles 50

Galicia

• Seeking independence from Spain

• Political parties: Galician Nationalist Bloc (separatist), Compromiso por Galicia (separatist)

Logo of the Bloque Nacionalista Galego (Galician Nationalist Bloc)

• Resources: Galicia’s resources are negligible in comparison to Catalonia’s resources. Its average share of Spain’s GDP between 2000 and 2010 was 5.2 percent, which is less than a third of Catalonia’s share (1).

• Electoral success and party strategies: Support for independence has been much less intense than in Catalonia, with the Galician Nationalist Bloc receiving only 10 to 20 percent of the vote. For this reason, Duerr(2016) classiﬁes Galicia’s case as a “second –tier” secessionist movement (2).

• Sources: (1) See Instituto Nacional de Estadística at http://www.ine.es/en/daco/daco42/cre00/b2010/ homog/dacocre_base2010h_en.htm, last accessed on 31st October 2016 (2) See Duerr(2016, p. 11)

• Quotes: “Although a direct comparison of Galicia and Catalonia is challenging, there are points of comparison that can be made. Support for BNG has fluctuated between 10-20 percent, but never has the party attained the level of CiU and ERC combined. Even though regional elections were held in different years, a comparison of voter support for independence shows that Galician pro-independence support has remained stagnant; whereas, support for Catalan secession has grown, especially with the aforementioned switch by CDC to pro-independence positions” (Duerr, 2016, p. 11) A Region Profiles 51

Greenland

• Seeking independence from Denmark

• Political parties: Inuit Ataqatigiit (separatist), Forward (separatist), Inuit Party (separatist)

Logo of the Inuit Ataqatigiit (Community of the People)

• Resources: The sparsely populated island (56,648 inhabitants) still strongly relies on the historic fish industry as the largest income earner (1), which does not generate enough revenue to finance Green- land’s public expenditures, wherefore a Danish grant of 3.6 billion kroner ($604m) accounts for over half of Greenland’s revenues (2). The development of mining (rare metals and radioactive substances since 2013 (3)) and oil industries (discoveries by Carin Energy in 2010 (4)) spurred independence movements (3), but falling crude prices rendered independence less financially viable, with recent studies estimating that Greenland will depend on Danish grants for at least another 25 years (2).

Electoral success and party strategies: The recent oil discoveries instilled hopes for ﬁnancial independence in separatist leaders, with the former prime minister Aleqa Hammond claiming that independence is possible “within her lifetime”. But falling crude prices have made the new prime minister unequivocally less optimistic, and studies estimate that Greenland will remain ﬁnancially dependent on Denmark for at least another 25 years (2).

• Sources: (1) Government of Greenland, Economy and Industry in Greenland at http://naalakkersuisut. gl/en/About-government-of-greenland/About-Greenland/Economy-and-Industry-in-Gre enland, last accessed on 31st October 2016 (2) See The Economist from 21st January 2015 at http://www.economist.com/news/europe/ 21640224-falling-crude-prices-are-forcing-greenland-put-plans-split-denmark-ind ependence-ice, last accessed on 31st October 2016 (3) The Economist from 31st March 2013 http://www.economist.com/blogs/newsbook/2013/ A Region Proﬁles 52

03/economist-explains-why-greenland-election-global-implications, last accessed on 31st October 2016 (4) See The Economist from 26th August 2010 at http://www.economist.com/node/16889623, last accessed on 31st October 2016

• Quotes: “When Cairn Energy, a British petrochemicals company, discovered traces of oil beneath Greenland’s territorial waters in 2010, it seemed the secessionists’ prayers had been answered. Oil and other minerals including aluminum and gold, it was hoped, would give the territory of just 56,200 inhabitants the ﬁnancial clout to go it alone” (The Economist from 21st January 2015 at http://www.economist.com/news/europe/21640224-f alling-crude-prices-are-forcing-greenland-put-plans-split-denmark-independence-i ce, last accessed on 31st October 2016)

“Greenland’s politicians were emboldened by the prospect of petrodollars. Aleqa Hammond, who served as her country’s ﬁrst female prime minister between April 2013 and September 2014 (when a corruption scandal drove her from oﬃce), said independence was possible “within her lifetime”. [...] One year later, the political rhetoric has dropped a few tones. At a press conference on January 9th in Copenhagen, the new prime minister, Kim Kielsen, said the “light of independence burned within” but he was unsure if it would be realised in his lifetime. Mr Kielsen is 48, suggesting that the timeline has been pushed back a few decades.” (The Economist from 21st January 2015 at http://www.economist.com/news/europe/21640224-f alling-crude-prices-are-forcing-greenland-put-plans-split-denmark-independence-i ce, last accessed on 31st October 2016)

“The world may not often be very interested in Greenland but it is fascinated by what lies beneath it. As the country’s ice cap melts, hidden mineral wealth is coming tantalisingly within reach. The country’s riches include “rare earth” metals that are essential in the production of many electronic devices, from electric-car batteries to television screens. Metals such as cerium (used in glass manufacturing) and yttrium (which goes into electronic displays) are among those that are hidden under the ice.” (See The Economist from 31st March 2013 http://www.economist.com/blogs/newsbook/2013/ 03/economist-explains-why-greenland-election-global-implications, last accessed on 31st October 2016) A Region Proﬁles 53

Québec

• Seeking independence from Canada

• Political parties: Parti Québécois (separatist), Bloc Québécois (separatist), Québec Solidaire (separatist), Option nationale (separatist), Parti indépendantiste (separatist), Marxist–Leninist Party of Quebec (separatist)

Logo of the Parti Québécois

• Resources: Québec’s share of the national GDP has decreased more or less steadily from 20.8 percent in 1993 to 19.4 percent in 2006. It then increased by 0.7 percent from 2006 to 2009 before falling back to 19.1 percent in 2014, the lowest value in the last 20 years (1). Supposedly almost $ 53 billion are transferred to the federal system, which are then redistributed across Canada (2).

Source: Government of Canada (Statistics Canada) at http://www.statcan.gc.ca/eng/start, last accessed on 31st October 2016 A Region Proﬁles 54

Electoral success and party strategies: The Parti Québécois argues that Québec could increase its budget by 70 percent by gaining independence, as the region annually transfers $52.9 billion to the federal system and supposedly gets back less in return (2). The collective vote share of the two biggest separatist parties on the territorial level Parti Québécois and Québéc Solidaire in the National Assembly of Québec has developed similarly. Parti Québécois held 77 seats, corresponding to 61.6 percent of the seats, after the 1994 Election, but the percentage declined to 28.8 percent in 2007. In 2008, when Québéc Solidaire won its ﬁrst seat in the Assembly, the two parties surged and won a combined 41.6 percent of the seats (3)

• Sources: (1) Government of Canada at http://www5.statcan.gc.ca/cansim/a26?lang=eng&retrLang= eng&id=3840038&&pattern=&stByVal=1&p1=1&p2=35&tabMode=dataTable&csid=, last accessed on 31st October 2016 (2) See http://mon.pq.org/documents/monpq_56e6f29687b17.pdf, last accessed: on 31st Octo- ber 2016 (3) See Le directeur général des élections du Québéc http://www.electionsquebec.qc.ca/fr ancais/provincial/resultats-electoraux/elections_generales_statistiques.php, last accessed on 31st October 2016

• Quotes: “Quebec is rich. It has the currently best administration to confirm this, but the federal government benefits every year from tens of billion sent by Quebec. To get back those $ 52.9 billion would increase A Region Profiles 55

Quebec’s budget by almost 70 percent, thanks to our economic development.” (Economic Arguments, Parti Québécois (p. 2) at http://mon.pq.org/documents/monpq_56e6f 29687b17.pdf, last accessed: on 31st October 2016)

New Caledonia

• Seeking independence from France

• Political parties: Kanak and Socialist National Liberation Front (Caledonian Union, Party of Kanak Liberation, separatist), Kanak Socialist Liberation (separatist)

Logo of the Kanak and Socialist National Liberation Front (Front de Libération Nationale Kanak et Socialiste)

• Resources: As a French colony since 1853, New Caledonia still strongly relies on ﬁnancial assistance from mainland France (1). This is hoped to be gradually alleviated through further expansions of the New Caledonian nickel industry, as the island is believed to hold roughly a quarter of the world’s nickel resources and currently ranks 5th among the top nickel-producing countries (2). The Koniambo Project, a nickel mine in which Xstrata (merged with Glencore in 2013) invested $6 billion, is the largest recent expansion of the Caledonian nickel industry (4, 5).

Electoral success/party strategies: Independence activists hope that achieving economic independence in the near future will lay the foundation for complete political independence. The Koniambo Project has further instilled hopes for independence in independence leaders, which emphasize the opportunity for a largely Kanak organization (Société minière du Sud Pacifique) to work with a non-French company (Xstrata, now Glencore Xstrata), further reducing New Caledonia’s economic dependence on mainland France. The fact that Société minière du Sud Pacifique, the involved local mining company, is owned by Kanak from the largely pro-independence Nothern Province is viewed as an additional benefit (3). A Region Profiles 56

• Sources: (1) See The Economist from 25th May 2013 at http://www.economist.com/news/asia/21578438-p ressures-independence-are-alive-not-always-kicking-ends-empire, last accessed on 31st Oc- tober 2016 (2) See, e.g. BBC News from June 16, 2016 at http://www.bbc.com/news/world-asia-pacific -16740838, last accessed on 31st October 2016 (3) See Horowitz(2004) (4) See Financial Times from 7th June 2007 at http://www.ft.com/cms/s/0/b5d6b672-1494-11d c-88cb-000b5df10621.html?ft_site=falcon&desktop=true#axzz4LXUJ59MK, last accessed on 31st October 2016 (5) Bloomberg from 2nd August 2011 at http://www.bloomberg.com/news/articles/2011-08-02/ xstrata-first-half-profit-rises-27-as-commodity-prices-climb, last accessed on 31st Oc- tober 2016

• Quotes: “In New Caledonia, pro-independence leaders perceive economic autonomy as a prerequisite for political independence. The Koniambo Project, a joint venture between a Canadian multinational and a local mining company, is seen by many Kanak as an opportunity to loosen economic ties to metropolitan France” (Horowitz, 2004, p. 318)

“For half-century pioneers developed the idea that the Caledonians had the right to decide what to do with their mineral resources. From there on, this concern was central to the commitment to independence of the FLNKS Front: to have control. To have control over our natural resources, to have control over industrial tools, to have the control over mining and metallurgical annuity.” (Statement on Nickel Mining, Caledonian Union at http://unioncaledonienne.com/wp-conte nt/uploads/2015/10/D%C3%A9claration-liminaire-UC-FLNKS-14-10-15.pdf, last accessed on 31st October 2016)

“The Paciﬁc territory with the most realistic chance of decolonisation is nickel-rich New Caledo- nia, a French colony since 1853.” (The Economist from 25th May 2013 at http://www.economist.com/news/asia/21578438-pre ssures-independence-are-alive-not-always-kicking-ends-empire, last accessed on 31st Oc- tober 2016) A Region Proﬁles 57

Republika Srpska

• Seeking independence from Bosnia and Herzegovina

• Political parties: Alliance of Independent Social Democrats (secessionist), Serbian Democratic Party (secessionist)

Logo of the Alliance of Independent Social Democrats

• Resources: The region possesses 20 percent of Bosnia and Herzegovina’s resources of brown coal and 40 percent of lignite among various other mineral resources (1). In 2012 Republika Srpska’s president reached an agreement with the Russian company Gazprom to build part of the South Stream pipeline network, reinforcing the region’s strategic advantage concerning the energy sector (2).

• Electoral success and party strategies: Prominent secessionist like the region’s current president Milo- rad Dodik (Alliance of Independent Social Democrats) argue that “only an independent RS could reach its full economic potential”, emphasizing the region’s economic strength.

• Sources: (1) See Steblez(2005) (2) See LSE Blog from 16th September 2014 at http://blogs.lse.ac.uk/lsee/2014/09/16/ga zproms-gas-deal-in-republika-srpska-provides-south-stream-stop-gap/, last accessed on 31st October 2016

• Quotes: “He [Milorad Dodik, president of Republika Srbska since 2010] called for another “extraordinary” RS Assembly to meet and organize a referendum in March 2010. “Sarajevo’s constant accusations of separatism against the RS could indeed lead to secession,” Dodik indicated. Presenting a paper called “Your Srpska, Your Vote,” he declared that sovereignty was the goal and that only an independent RS could reach its full economic potential” (Toal, 2013, p. 192). A Region Proﬁles 58

Santa Cruz

• Seeking independence from Bolivia

• Political parties: Pro Santa Cruz Committee (separatist)

Logo of the Pro Santa Cruz Committee (Comite Pro Santa Cruz)

• Resources: Santa Cruz is considered one of South America’s most important energy hubs, and is expected to gain even more relevance in the future (1). The region is believed to possess the second largest natural gas deposit in South America (2), and gas exports increased ﬁve-fold from 2000 to 2005 to $620m. In addition, Santa Cruz has the highest regional GDP in Bolivia and its capital city is widely recognized as Bolivia’s economic capital (3).

Electoral success and party strategies: In 2008 over 85 percent voted in favor of autonomy in an unoﬃcial referendum which was later declared unconstitutional. Nonetheless, the referendum bol- stered up Santa Cruz’s leaders pushing for more autonomy to gain control over land, oil and taxes, which they feel are currently unfairly distributed by Bolivia’s system of regional redistribution (3, 4).

• Sources: (1) See The Economist 21st April 2005 at http://www.economist.com/node/3896009, last accessed on 31st October 2016 (2) See Forbes Custom from 28th September 2015 at http://www.forbescustom.com/SectionPD Fs/092815-Bolivia-EcDev.pdf, last accessed on 31st October 2016 (3) See The Economist from 8th May 2008 at http://www.economist.com/node/11332954, last accessed on 31st October 2016 (4) See the Financial Times from 8th May 2008 at https://www.ft.com/content/0cbdb74a-19f f-11dd-ba02-0000779fd2ac, last accessed on 31st October 2016

• Quotes: “The capital of Bolivia’s easternmost province, Santa Cruz, is prosperous and seething with frustration. A Region Proﬁles 59

It is a centre of commercial farming, and the local headquarters for multinationals pumping natural gas out of the country. There is nothing cruzeños want more than to export gas and soya beans. But for much of the past two years, radical groups from Bolivia’s western highlands have done their best to prevent that. [...] To inoculate itself against such disruption, Santa Cruz wants to gain autonomy from the central government in La Paz.” (The Economist 21st April 2005 at http://www.economist.com/node/3896009, last accessed on 31st October 2016)

“Santa Cruz is the department with the most prosperous and dynamic economy in Bolivia. Of- ficial statistics confirm that it has established itself as the region which contributes the most to the national GDP and as the department which contributes a large part of the national tax collection, produced by the most productive, diversified and modern economy of the country.” (Development Model for Santa Cruz 2014, Santa Cruz Committee (p. 4) http://www.comitepros antacruz.org.bo/imagen/documentos/2.pdf, last accessed on 31st October 2016)

Istria

• Seeking independence from Croatia

• Political parties: Istrian Democratic Assembly (separatist), Istrian Democratic Forum (separatist)

Logo of the Istrian Democratic Assembly (Istarski demokratski sabor)

• Resources: Istria is considered the most important tourist destination in Croatia, realizing 27 percent of all arrivals and 35 percent of all nights spent in the country. In addition, the region features a highly developed processing and shipping industry (1).

• Electoral success and party strategies: Between 1992 and 2015, the Istrian Democratic Assembly contin- uously held seats in the Hrvatski Sabor, the Croatian Parliament (2). A Region Proﬁles 60

• Sources: (1) See Government of Istria, Istrian Economy at https://www.istra-istria.hr/index.php?i d=603, last accessed on 31st October 2016 (2) See Interparliamentary Union at http://www.ipu.org/parline-e/reports/2077_arc.htm, last accessed on 31st October 2016

• Quotes: “Like most subnational regional movements, Istrian regionalism was not motivated primarily by ethnic or national belonging. Such movements, like the economic regionalism in Germany in the 1980s, are driven more by economic concerns and the drive for subsidiarity than by cultural or ethnic criteria.” (Ashbrook & Ashbrook, 2008, p. 151)

Upper Silesia

• Seeking independence from Poland

• Political parties: Silesian Autonomy Movement (separatist)

Logo of the Silesian Autonomy Movement (Ruch Autonomii Śląska)

• Resources: The region possesses extensive lignite and brown coal deposits, with the state-owned Kompania Weglowa (KW) being the largest coal-mining company in Europe (1) and 100,000 people employed in mines. In addition, the region features a ﬂourishing car manufacturing industry, large chemical works and leading scientiﬁc research institutions, together make Upper Silesia the second richest of Poland’s 16 voivodships (2).

• Electoral success and party strategies: In 2010 the Silesian Autonomy Movement election slogan was “Silesian Money for Silesian People”, very reminiscent of the SNP’s “It’s Scotland’s Oil!” campaign. The election campaign was centered on Poland’s system of regional redistribution, which separatist leaders argue takes too much from Upper Silesia’s tax money to distribute elsewhere. At the election A Region Proﬁles 61

for the district parliament in 2010 the Silesian Autonomy Movement received 8.5 percent of the votes (3).

• Sources: (1) See The Economist from 28th June 2014 at http://www.economist.com/news/special-repor t/21604686-traditional-industries-are-declining-outsourcing-offshoring-and-sub contracting-are, last accessed on 31st October 2016 (2) See The Guardian from 8th April 2011 at https://www.theguardian.com/world/2011/apr/ 08/upper-silesia-flags-up-independence, last accessed on 31st October 2016 (3) See Bundeszentrale für politische Bildung http://www.bpb.de/internationales/europa/po len/202995/tabellen-und-grafiken-zum-text-wahlergebnisse-in-der-woiwodschaft-s chlesien, last accessed on 31st October 2016

• Quotes: “But whereas Scotland has drilled down into the North Sea to make the money it resents being made to channel via Westminster, Upper Silesia’s riches come from under solid ground. It still employs 100,000 people in coalmines, and thousands more in the many steelworks. Plus, it boasts a booming car manufacturing industry – Opel has a plant in Gliwice and Fiats are made in Tychy and Bielsko-Biala – and big chemical works at Kedzierzyn Kozle and Zdzieszowice, and has a great track record for scientific research, particularly in clean coal technology, soil detoxification and renewable energy. "We are officially the second richest of 16 voivodships in Poland, after Warsaw and Masovia, and provide 14 percent of the GDP," said Gorzelik [leader of the RAS], "and we feel we don’t get enough back from the national government." The RAS’s election slogan last year was "Silesian Money for Silesian People", arguing that Upper Silesia should receive more money back from Warsaw, and be given the autonomy to spend it as it wishes.”

“The money, which will develop the people of our region, will remain at our disposal. The inhabitants of the land will decide on the distribution of these funds. The Silesian mining law will be discussed by local experts from the mining industry, not the MPs from Szczecin.” (FAQ Section, Silesian Autonomy Movement at http://autonomia.pl/faq/, last accessed on 31st October 2016) A Region Proﬁles 62

Northern Italy

• Seeking Independence from Italy

• Political parties: Lega Nord per l’Indipendenza della Padania (formerly secessionist, separatist since 2006)

Logo of the Lega Nord per l’Indipendenza della Padania (North League)

• Resources: Northern and Southern Italy are regularly referenced as a particularly salient example of regional economic divide, owing partially to its longevity. Northern Italy was the ﬁrst part of Italy to industrialize in the 19th century, and remains the most developed and productive area of Italy to the present day. For instance, unemployment in 2014 was 21.7 percent in Southern Italy, compared with only 13.6 percent in the whole of Italy, indicating not only a much weaker economic performance, but also a lack of human capital in comparison to Northern Italy (1).

• Electoral success and party strategies: Unlike most other regionalist parties, the Lega Nord could not rely on a historic nation-state as an argument for independence, but instead proclaimed it’s own hypothetical state called “Padania”. In the Lega Nord’s Padanian Declaration of Independence from 1996, the economic strength of the region is put forward as a key argument for independence, while the Italian central state is accused of economically exploiting “Padania”. The results of the Lega Nord in the Chamber of deputies after the introduction of the new electoral system in 2005 ﬂuctuated. They held 26 of the 617 seats after the 2006 Election and even increased its share to 60 seats in 2008. It has decreased signiﬁcantly in 2013, when Lega Nord lost a total of 42 seats (2).

• Sources: (1) See The Economist from 16th May 2015 at http://www.economist.com/news/finance-and-e conomics/21651261-north-limps-ahead-south-swoons-tale-two-economies, last accessed on 31st October 2016 (2) See Election Resources at http://www.electionresources.org/it/, last accessed 31st Octo- ber 2016 A Region Proﬁles 63

• Quotes: “In contrast, the history of the Italian State has become the history of colonial oppression, of economic exploitation, and of moral violence; The Italian State has, over time, systematically occupied Padania’s economic and social system through its parasitic bureaucratic apparatus.” (Padanian Declaration of Independence 1996 at https://web.archive.org/web/20001207094000/ http://www.leganord.org/frames/english.htm, last accessed on 31st October 2016)

“The key diﬀerence between the LN’s political project and the majority of other regionalist political parties is the fact that it is not based in an area that has historic claims to nationhood. Instead, the LN has attempted to invent an ethnicity for the North of Italy in order to justify its political claims for the protection of the economic interests of the region.” (Giordano, 2000, p. 446) B Giant Discoveries 64

B Giant Discoveries

Name Year Discovered Start of Production Reserves Alba 1984 Jan. 1994 400 MMstb. Alwyn North 1971 Nov. 1987 309 MMstb. Andrew 1974 Jun. 1996 150 MMstb. Arbroath 1969 Apr. 1990 97.9 MMstb. Balmoral 1975 Nov. 1986 100 MMstb. Beatrice 1979 Sep. 1981 495 MMstb. Bentley 1977 Jul. 1905 880.9 MMstb. Beryl 1972 Jun. 1976 2100 MMstb. Brae-North and South 1975 Jun. 1905 70 MMstb. Brent 1971 Nov. 1976 3500 MMstb. Bressay 1978 Jul. 1905 200 MMstb. Buchan 1974 May 1981 120 MMstb. Buzzard 2001 Jan. 2007 1500 MMstb. Captain 1977 Mar. 1997 700 MMstb. Clair 1977 Feb. 2005 5000 MMstb. Claymore 1974 Nov. 1977 662 MMstb. Cormorant North 1972 Feb. 1982 90 MMstb. Crawford 1975 Apr. 1989 130 MMstb. Donan 1987 Jan. 2007 60.3 MMstb. Douglas (Wales) 1990 Jan. 1996 225 MMstb. Dunbar (Alwyn S. S. A.) 1972 Dec. 1994 850 MMstb. Dunlin 1973 Aug. 1978 363 MMstb. Eider 1976 Nov. 1988 85 MMstb. Elgin-Franklin Fields 1985 Jun. 1905 365 MMstb. ETAP 1995 Nov. 1998 490 MMstb. Foinaven 1990 Nov. 1997 600 MMstb. Forties 1970 Sep. 1975 5000 MMstb. Fulmar 1975 Feb. 1982 73 MMstb. Gannet (A,C,D,E,F,G) 1973 Nov. 1993 214 MMstb. Golden Eagle 2001 Nov. 2014 140 MMstb. B Giant Discoveries 65

Name Year Discovered Start of Production Reserves Gryphon 1987 Oct. 1993 207 MMstb. Harding 1987 Apr. 1996 322 MMstb. Heather 1973 Oct. 1978 464 MMstb. Hutton 1973 Aug. 1984 265 MMstb. Ivanhoe 1975 Jul. 1989 100 MMstb. Janice 1990 Feb. 1999 70 MMstb. Kittiwake 1981 Sep. 1990 70 MMstb. Kraken 1985 Jul. 1905 137 MMstb. Leadon 1989 Jun. 1905 120 MMstb. Macculloch 1990 Aug. 1997 58 MMstb. Magnus 1974 Aug. 1983 1540 MMstb. Mariner Oilﬁeld 1981 Jul. 1905 250 MMstb. Maureen 1973 Sep. 1983 210 MMstb. Miller 1983 Jun. 1992 345 MMstb. Montrose 1971 Jun. 1976 93.6 MMstb. Murchison (UK) 1975 Sep. 1980 400 MMstb. Nelson 1988 Feb. 1994 790 MMstb. Ninian 1974 Dec. 1978 2920 MMstb. Northwest Hutton 1975 Jun. 1905 265 MMstb. Osprey 1974 Jun. 1905 158 MMstb. Pierce 1975 Feb. 1999 100 MMstb. Piper 1973 Dec. 1976 618 MMstb. Ross 1981 Apr. 1999 100 MMstb. Saltire 1988 May 1993 224 MMstb. Schiehallion 1993 Jun. 1905 450-600 MMstb. Scott 1983 Sep. 1993 440 MMstb. South Cormorant 1972 Dec. 1979 90 MMstb. Tartan 1974 Jan. 1981 116 MMstb. T-Block 1976 Nov. 1993 100 MMstb. Tern 1975 Jun. 1989 175 MMstb. Thistle 1972 Feb. 1978 824 MMstb.

A main data source was https://www.gov.uk/guidance/oil-and-gas-uk-field-dataZuk -oil-and-gas-reserves-and-resources, last accessed on 31st October 2016. B Giant Discoveries 66

Each individual discovery, its discovery date, and size were veriﬁed using various sources. These were: Casey et al. (1993); Coward et al. (1991); Eneyok et al. (2003); EnQuest(2013, n.d.); E.ON(2013); Favero et al. (1994); Fee & O’Dea(2005); Glennie & Armstrong(1991); Guscott et al. (2003); Jayasekera et al. (1999); Kavanagh(2013); Kay(2003); Kunka et al. (2003); Nexen/CNOOC(2013); Pye & Brown(2002); Ritchie(2003); Talisman Energy(2006a,b, 2007); The Maersk Group(2014); Tonkin & Fraser(1991); United Kingdom Government(2013); Van Vessem & Gan(1991); Walker(1994). C Calculation of Constituency Results 67

C Calculation of Constituency Results

To compute the election results for the individual constituencies based on the 2001 boundaries, we applied the following procedure. First, we superimposed historical electoral maps with a graphical software to detect whether constituency boundaries have changed. For each period, we calculated how the 2001-constituencies consist of the historical constituencies used in previous GEs. As there exists no better estimate for the population distribution within a constituency, we assume a uniform distribution. To describe the overlapping area of the historical constituencies and the 2001-constituencies, we use ﬁfths gradations. Election results of constituencies included in the panel are then extrapolated based on the following formula: Assume the historical constituencies 1, 2,.., n from the GE in t overlap with the 2001-constituency j. The extrapolation Pn x , × Y , = i=1 i t i t ∈ { , 1, 2, 3, 4, } for this constituency is given by: Yj,t Pn , xi,t 0 5 5 5 5 1 , with Yi,t being the election i=1 xi,t result of an original constituency in year t, Yj,t being the projected result of a 2001-constituency in t, and xi,t representing the shares of the n original constituencies i overlapping with the 2001-constituency j. By this pattern, the results from historical constituencies are included in a weighted form in the extrapolated result based on the boundaries of 2001-constituencies. If only one historical constituency accounts for a 2001-constituency, the result is adopted without further computation. For transparency reasons, we provide the entire weightings for all constituencies in all time periods on the following pages.

The map shows the Scottish 2001-constituences to which the older election results are projected. It provides an example of the GIS maps, which we used to build the dataset analysed. Source: http://www.bcomm-scotland.in dependent.gov.uk/maps/datafiles/. C Calculation of Constituency Results 68

Sources: Boundary Commission for Scotland (http://www.bcomm-scotland.independent.gov.uk/m aps/datafiles/), David Boothroyd (http://www.election.demon.co.uk/), UK Data Service (https: //census.edina.ac.uk/easy_download.html); all last accessed on 19th August 2015. C Calculation of Constituency Results 69 Kincardeene and Deeside Angus East 3/5 Provan Banff and Buchan Bridgend + 2/5 Gordon + 3/5 Motherwell North + 3/5 Shettleston 3/5 West Aberdeenshire Mearns 2/5 South Angus Mearns 3/5 Hillhead + Garscadden 1/5 Hillhead + Garscadden Ayr3/5 East Aberdeenshire Ayr 1/5 Aberavon + 3/5 North Angus and + 1/5 North Angus and + 1/5 Ruhterglen + 2/5 Banff + 2/5 Ogmore 3/5 West Aberdeenshire Mearns 2/5 South Angus Mearns 3/5 Hillhead 4/5 Ayr + 1/5 Central 3/5 Provan +1/5 Bothwell 3/53/5 Provan East Aberdeenshire 1/5 Aberavon Ayrshire + 3/5 North Angus and + 1/5 North Angus and + 4/5 Scotstoun + 2/5 Banff + 2/5 Ogmore 3/5 West Aberdeenshire Mearns Aberdeen North Aberdeen North Aberdeen North2/5 South Angus Mearns Aberdeen North 3/5 East Aberdeenshire 4/5 Ayr + 1/5 Central 1/5 Aberavon Caernarvon Caernarvon Caernarvon Caernarvon Ayrshire + 3/5 North Angus and + 1/5 North Angus and + 4/5 Scotstoun + 2/5 Banff + 1/5 Bothwell + 2/5 Ogmore Kincardine and West District of Burghs Caithness and Sutherland Caithness and Sutherland Caithness and Sutherland Caithness and Sutherland Caithness and Sutherland Aberdeenshire + 2/5 Aberdeen South + 2/5 Banff + 1/5 Kilmarnock + 2/5 Ogmore + 2/5 Caernarvon District Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 Kincardine Easter Ross Scotland AberdeenScotland North AberdeenScotland South Aberdeenshire West and CentralScotland Aberdeenshire Airdrie and Aberdeen Shotts SouthWales West AberdeenshireScotland Alyn and Deeside Angus North Lanarkshire Aberdeen South West Aberdeenshire Flint North Lanarkshire West Aberdeenshire Aberdeen South 2/5 Forfar +2/5 Montrose North Lanarkshire Aberdeen North Aberdeen South East Flint North Lanarkshire Aberdeen South Monklandes East East Flint East Flint Alyn and Deeside WalesScotland Aberavon Aberdeen Central Aberdeen North Aberavon Aberavon Aberavon Aberavon Aberavon Scotland Anniesland (Glasgow) 2/5 Hillhead + 2/5 Pattrick 3/5 Hillhead Scotland Banff and Buchan 3/5 East Aberdeenshire Scotland ArgyllScotland and Bute AyrScotland Baillieston (Glasgow) Argyll Bothwell 2/5 Ayr District of Burghs Argyll 3/5 Camlachie Argyll Argyll Argyll and Bute WalesWales Blaenau GwentWales Brecon and Radnor Bridgend Ebbw Vale Brecon +2/5 and Abertillerie Rednor Ebbw Vale +2/5 Abertillerie Ebbw Vale +2/5 Abertillerie 1/5 Aberavon Brecon and Rednor Ebbw Vale +2/5 Abertillerie Blaenau Gwent Brecon and Rednor Brecon and Rednor Brecon and Rednor Wales Cardiff Central Cardiff Central Cardiff North Cardiff North Cardiff North Cardiff Central Wales CaernarvonWalesScotland Caerphilly Caithness, Sutherland and Caernarvonshire Caerphilly Caerphilly Caerphilly Caerphilly Caerphilly C Calculation of Constituency Results 70 North 1/5 Camarthen 3/5 Wrexham + 2/5 Clwyd South West Monklands West Delyn Valley + 2/5 Pembroke + 2/5 Clwyd North West 1/5 Camarthen 3/5 Wrexham Central Dunbartonshire Clydebank and Milngavie 1/5 North Lanarkshire 4/5 Flint West + 2/5 Flint East Airdrie + 2/5 Pembroke + 2/5 Denbigh + 3/5 Coatbridge and 1/5 Camarthen 3/5 Wrexham 1/5 East Dunbartonshire Dunbartonshire 1/5 North Lanarkshire 4/5 Flint West + 2/5 Flint East Airdrie + 2/5 Pembroke + 2/5 Denbigh + 3/5 Central + 3/5 Coatbridge and Cardiff North + 1/5 Barry Cardiff North + 1/5 Barry Cardiff North West1/5 Camarthen Cardiff North 3/5 Wrexham 1/5 East Dunbartonshire Dunbartonshire 1/5 North Lanarkshire ConwayCentral Ayrshire Conway Central AyrshireEast Conway Central Ayrshire Cunninghame South Conway Airdrie + 2/5 Pembroke + 2/5 Denbigh + 3/5 Central + 3/5 Coatbridge and Carmarthen1/5 Camarthen Carmarthen Carmarthen Carmarthen Carmarthen South Ayrshire South Ayrshire South Ayrshire South Ayrshire Carrick, Cunnock and Doon Burghs bridge+ 1/5 Bothwell Ayrshire + 2/5 Pembroke + 2/5 Denbigh + 1/5 Llandaff and Barry + 2/5 Dumbarton District of + 3/5 Coath- + 3/5 Caernarvon District + 1/5 Bute and North Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 Pembrokeshire South Dinefwr Doon Valley Wales Carmarthen West and WalesWales Cardiff South and PenarthWales Cardiff West Carmarthen Cardiff East South + and Cardiff East Cardiff South East Llandaff and BarryScotland Cathcart CardiffWales (Glasgow) South East Ceredigion Gogledd PenfroWales Cardiff West Cardigan 1/5 Rutherglen Clwyd + South Cathcart Cardiff South East Cathcart Cardiff West Cardiff South and Penarth 3/5 Wrexham Cardigan Cathcart Cardiff West Cardigan Cathcart + 1/5 Pollok Cardiff West Cathcart Cardigan Ceredigion and Pembroke Wales Cardiff North Cardiff North Scotland Carrick, Cumnock and Wales Clwyd WestScotland Clydebank and Milngavie Dunbartonshire Denbigh Denbigh Denbigh Denbigh 2/5 Clwyd South West Scotland ClydesdaleScotland Coatbridge and Chryston 1/5 North Lanarkshire Lanark Lanark Lanark Lanark Clydesdale Wales ConwayScotland CumbernauldScotland and Kilsyth CunninghameScotland North East Cunninghame Dunbartonshire South Bute andWales North Ayrshire Caernarvonshire Dunbartonshire 1/5Wales Ayr District of Cynon Burghs Valley Bute and North Delyn Ayrshire Bute and North Ayrshire Dunbartonshire Aberdare + 1/5 Merthyr Bute and North Ayrshire Aberdare Cunninghame Flint East North Dunbartonshire Cumbernauld und Kilsyth Aberdare 4/5 Flint West + 2/5 Flint Aberdare Cynon Valley C Calculation of Constituency Results 71 Dunfermline East Edinburgh Central Edinburgh Leith Edinburgh West Falkirk East Falkirk West + 1/5 Dunfermline West + 1/5 Edinburgh West + 1/5 Livingston + 1/5 Linlithgow + 1/5 Clackmannan Dunfermline Dunfermline West Edinburgh Leith Edinburgh West +1/5 West Lothian 3/5 Stirling and Falkirk Lothian + 3/5 Clackmannan and East Stirling Stirling and Falkirk Grangemouth + 1/5 West Grangemouth + 2/5 Dunfermline + 1/5 Edinburgh North + 1/5 Edinburgh West + 3/5 Edinburgh North 2/5 West Fife Edinburgh Leith Edinburgh West +1/5 West Lothian 3/5 Stirling and Falkirk Lothian + 3/5 Clackmannan and East Stirling Stirling and Falkirk Grangemouth + 1/5 West Grangemouth + Dunfermline Burghs + 3/5 Edinburgh North West Dunbartonshire West Dunbartonshire West Dunbartonshire2/5 West Fife Dumbarton Edinburgh Central Edinburgh Central Edinburgh Central Edinburgh Leith Edinburgh West + 1/5 West Lothian 3/5 Stirling and Falkirk Lothian + 3/5 Clackmannan and East Stirling Stirling and Falkirk Grangemouth + 1/5 West Grangemouth + Dunfermline Burghs + Edinburgh North Burghs ton District of Burghs Edinburgh East Edinburgh East Edinburgh East Edinburgh East Edinburgh East Stirling Stirlingshire + 1/5 Stirling Districts of Burghs + Dunfermline District of + Dumbar- + Edinburgh Central + 1/5 Edinburgh West + 1/5 Linlithgowshire +1/5 Linlithgowshire Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 Musselburgh Scotland DumfriesScotland DundeeScotland East DundeeScotland West Dunfermline EastScotland Dunfermline West Dumfriesshire Dundee West Dundee Fife 2/5 West Fife Dumfries Dundee East West Fife Dundee West Dumfries Dundee East Dundee West West Fife Dumfries Dundee East Dundee West 3/5 Central Fife Dumfries Dundee East Dundee West Scotland Dumbarton 4/5 Dunbartonshire Scotland EastScotland Kilbride EastScotland Lothian EastwoodScotland Edinburgh Central Lanark Berwick and Haddingon Edinburgh West East Renfrewshire Berwick and East Lothian Berwick and East Lothian Lanark East Renfrewshire Berwick and East Lothian East Lothian East Renfrewshire Lanark East Renfrewshire Eastwood East Kilbride East Kilbride Scotland Edinburgh North and Leith Leith + Edinburgh North Scotland Edinburgh East and Scotland EdinburghScotland South Edinburgh West Edinburgh South 1/5 North Midlothian Edinburgh South Edinburgh South Edinburgh South Edinburgh South Scotland Falkirk East 2/5 Clackmannan and East Scotland Falkirk WestScotland Fife Central 1/5 Clackmannan and West 3/5 West Fife West Fife West Fife Central Fife Central Fife C Calculation of Constituency Results 72 Nithsdale 1/5 Banff und Buchan 1/5 Renfrew West and Inverclyde + 3/5 Greenock and Port Glasgow 2/5 Motherwell North Hamilton Inverness, Nairn und Lochaber 4/5 Hillhead +1/5 Central Livingston + 3/5 Gordon +1/5 Moray + 1/5 Central + 1/5 Hamilton 1/5 West Aberdeenshire 3/5 Greenock + 1/5 West Renfrewshire 3/5 Motherwell North 3/5 Inverness + 1/5 Argyll 2/5 Hillhead KilmarnockKirkcaldy Kilmarnock and Loudon 1/5 Midlothian Kirkcaldy Aberdeenshire + 1/5 Banff + 2/5 East + 2/5 Hamilton + 3/5 Hamilton + 2/5 Moray and Nairn + 3/5 Kelvingrove + 2/5 Central + 1/5 Westlothian 1/5 West Aberdeenshire Govan + 2/5 PollakGreenock + 1/5 West Govan + 1/5 PollakRenfrewshire 2/5 Govan + 1/5 Pollak 4/5 Bothwell 3/5 Inverness + 1/5 Argyll Kelvingrove side+ 2/5 Hillhead Kilmarnock + 1/5 Central 1/5 West Fife Aberdeenshire Ayrshire + 1/5 Banff + 2/5 East + 1/5 Hamilton + 2/5 Moray and Nairn + 2/5 Central+ 2/5 Wood- + 3/5 Kirkcaldy Burghs + 1/5 Westlothian 1/5 West Aberdeenshire 1/5 Pollok + 2/5 Govan Greenock + 1/5 West Renfrewshire 4/5 Bothwell Hamilton3/5 Inverness + 1/5 Argyll Hamilton 1/5 East Kilbride 2/5 Hillhead Kilmarnock + 1/5 Central 1/5 West Fife West Lothian 1/5 Midlothian 3/5 Woodside+ Maryhill 1/5 Woodside+ Maryhill 1/5 Kelvingrove+ Maryhill 1/5 Springburn+ Maryhill Aberdeenshire Ayrshire + 1/5 Banff + 3/5 East + Tradeston + 1/5 Hamilton + 2/5 Moray and Nairn + 2/5 Scoutstoun + Kelvin + 4/5 Central + 3/5 Kirkcaldy Burghs Galloway + 2/5 Dumfries Galloway + 2/5 Dumfries Galloway + 2/5 Dumfries Galloway + 2/5 Dumfries Galloway and Unpper Renfrewshire 4/5 Bothwell 3/5 Inverness + 1/5 Argyll and North Ayrshire Burghs Aberdeenshire + 1/5 Banff + 1/5 East + Tradeston + 1/5 Hamilton + 2/5 Moray and Nairn + 1/5 Rutherglen +Kelvingrove + 2/5 Hillhead + 2/5 Kirkcaldy District of + 1/5 Linlithgowshire + Maryhill Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 Nithsdale Bellshill Lochaber Scotland FifeScotland North East Galloway and Upper Fife East Fife East Fife East Fife East Fife North East Scotland Gordon 4/5 Central Aberdeenshire Scotland Govan (Glasgow)WalesScotland Gower Greenock and Inverclyde 4/5 Pollok + Govan Greenock + 1/5 West Gower Gower Gower Gower Gower Scotland Hamilton North and Scotland Hamilton SouthScotland Inverness East, Nairn and 3/5 Hamilton WalesScotland Islwyn Kelvin (Glasgow) 2/5 Patrick +4/5 Central Bedwellty Bedwellty + 2/5 Abertillery Bedwellty + 2/5 Abertillery Bedwellty + 2/5 Abertillery Islwyn Scotland Kilmarnock and Loudoun 3/5 Kilmarnock + 1/5 Bute Scotland KirkcaldyScotland LinlithgowScotland Livingston 1/5 West Fife Linlithgowshire 2/5 North Midlothian West Lothian West Lothian West Lothian Linlithgow WalesScotland Llanelly Maryhill (Glasgow) 2/5 Glasgow St. Rollox Llanelly Llanelly Llanelly Llanelly Llanelly C Calculation of Constituency Results 73 Meirionnydd Nant Conwy Rhymney Moray Newport East 4/5 Clackmannan Ogmore Perth and Kinross + 1/5 Perth and Kinross 4/5 Merionethshire 3/5 Moray and Nairn 2/5 Newport 1/5 Kinross and Stirlingshire 3/5 Ogmore 1/5 Kinross and and East Perthshire Westperthshire Westperthshire + 3/5 Perth (Merioneth) + 3/5 Conway +3/5 Banff + 3/5 Monmouth + 4/5 Clackmann and East + 1/5 Pontypridd + 1/5 Aberavon 4/5 Merionethshire 3/5 Moray and Nairn 2/5 Newport 1/5 Kinross and Stirlingshire 3/5 Ogmore 1/5 Kinross and and East Perthshire Westperthshire Westperthshire + 3/5 Perth (Merioneth) + 3/5 Conway +3/5 Banff + 3/5 Monmouth + 4/5 Clackmann and East + 1/5 Pontypridd + 1/5 Aberavon 4/5 Merionethshire Midlothian and Peebles Midlothian3/5 Moray and Nairn Midlothian Midlothian 2/5 Newport 1/5 Kinross and Stirlingshire 3/5 Ogmore PaisleyPaisley1/5 Kinross and Paisleyand East Perthshire Paisley Paisley Paisley 3/5 Paisley North Paisley South Westperthshire Westperthshire + 3/5 Perth (Merioneth) + 3/5 Conway +3/5 Banff + 3/5 Monmouth + 4/5 Clackmann and East + 1/5 Pontypridd + 1/5 Aberavon Merthyr +1/5 Caerphilly Merthyr +1/5 Caerphilly Merthyr +1/5 Caerphilly Merthyr +1/5 Caerphilly Merthyr Tydfil and Midlothian Stirlingshire Renfrewshire Perthshire Westperthshire (Merioneth) +3/5 Banff + 1/5 Caernarvonshire + 2/5 Caernarvon District + 1/5 Monmouth + 4/5 Clackmann and East + 1/5 Pontypridd + 1/5 Aberavon + 1/5 Paisley + 3/5 Perth and East Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 Rhymney Wales Merthyr Tydfil and Scotland MidlothianWalesWales MonmouthScotland Montgomeryshire Moray Peebles and South Montgomery Monmouth 3/5 Moray and Nairn Montgomery Monmouth Montgomery Monmouth Montgomery Monmouth Montgomery Monmouth Wales Meirionnydd Nant Conwy 4/5 Merionethshire Scotland MotherwellWales and WishawWales Neath Motherwell Newport East 2/5 Motherwell Newport Neath Motherwell 4/5 Neath + 1/5 Gower 4/5 Neath + 1/5 Motherwell Gower and Wishaw 4/5 Neath Motherwell + South 1/5 Gower Neath WalesScotland Newport West Ochil 2/5 Newport 1/5 Kinross and 2/5 Newport 2/5 Newport 2/5 Newport Newport West Wales Ogmore 3/5 Ogmore Scotland OrkneyScotland and Shetland Paisley NorthScotland Paisley Orkney South and ZetlandScotland PentlandsScotland (Edinburgh) 3/5 Paisley Perth + 1/5 East Orkney and Zetland North Midlothian 1/5 East Renfrewshire Orkney and Zetland Pentlands 1/5 Kinross and Orkney West... and Zetland Orkney and Shetland Pentlands Pentlands Pentlands C Calculation of Constituency Results 74 Pembroke North Ross, Skye and Inverness Roxburgh and Berwickshire 3/5 Central Stirling Strathkelvin and Bearsden North Tayside + 2/5 Angus East + 3/5 Pembroke + 2/5 Shetteston + 1/5 Provan + 2/5 Monklands West Rhondda2/5 Inverness + 3/5 Ross und Cromarty Rhondda 2/5 Berwick and East Lothian + 2/5 Roxburgh, Selkirk and Peebles 3/5 Queen’s Park 4/5 West Stirlingshire Perthshire 1/5 West Stirlingshire 2/5 Kinross and West Perthshire + 3/5 Perth an East P.+ 3/5 South Angus Mearns + 3/5 Central +Shettleston + 2/5 Kinross and West + 1/5 Dunbartonshire + 1/5 Noth Angus and 2/5 Inverness + 3/5 Ross und Cromarty Rhondda East + Rhondda 2/5 Berwick and East Lothian + 2/5 Roxburgh, Selkirk and Peebles 4/5 Gorbals + Bridgeton 4/5 West Stirlingshire Perthshire 1/5 West Stirlingshire 2/5 Kinross and West Perthshire + 3/5 Perth an East P.+ 3/5 South Angus Mearns West + Shettleston + 2/5 Kinross and West + 1/5 Dunbartonshire + 1/5 Noth Angus and 2/5 Inverness + 3/5 Ross und Cromarty 2/5 Pollok + 2/5 Govan 2/5 Pollok + Craigton 2/5 Pollok + CraigtonRhondda East + Rhondda Pollok + 2/5 Govan 2/5 Berwick and East Lothian + 2/5 Roxburgh and Selkirk Gorbals + Bridgeton 4/5 West Stirlingshire Perthshire 1/5 West Stirlingshire 2/5 Kinross and West Perthshire + 3/5 Perth an East P.+ 3/5 South Angus Mearns West + Shettleston + 2/5 Kinross and West + 1/5 Dunbartonshire + 1/5 Noth Angus and 2/5 Inverness + 3/5 Ross und Cromarty Renfrewshire Perthshire Perthshire + 4/5 Perth West + 2/5 Roxburgh and Selkirk + Shettleston + 2/5 Kinross and West + 1/5 Dunbartonshire +3/5 Forfar Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 West Scotland Ross, Skye and Inverness WalesWales Pontypridd Preseli PembrokeshireScotland RenfrewshireWales West Pembroke Rhondda Pontypridd West RenfrewshireScotland Roxburgh and Berwickshire Pembroke 2/5 Berwick and Haddington Pontypridd West Renfrewshire Rhondda East + Rhondda West Renfrewshire Pembroke Pontypridd Renfrew West and Inverclyde Pembroke Pontypridd 1/5 Ceredigion and Pontypridd Scotland Pollok (Glasgow) 1/5 Pollok + 1/5 East Scotland RutherglenScotland (Glasgow) Shettleston (Glasgow) Rutherglen Gorbals + Bridgeton Rutherglen Rutherglen Rutherglen Rutherglen Scotland Springburn (Glasgow)Scotland Stirling Springburn 1/5 Camlachie 4/5 West Stirlingshire Springburn + 2/5 Provan Springburn +1/5 Provan 4/5 Springburn Scotland Strathkelvin and Bearsden 1/5 West Stirlingshire WalesWales Swansea EastScotland Swansea West Tayside North Swansea East Swansea West 3/5 Kinross and West Swansea East Swansea West Swansea East Swansea West Swansea East Swansea West Swansea Swansea West East C Calculation of Constituency Results 75 Tewwdale, Ettrick and Lauderdale 3/5 Clwyd North West + 1/5 Clwyd South West 3/5 Roxburgh, Selkirk and Peebles + 1/5 Midlothian 3/5 Denbigh + 2/5 Flint West 3/5 Roxburgh, Selkirk and Peebles + 1/5 Midlothian Lothian 1/5 Denbigh + 2/5 Flint West + 1/5 Berwick and East Pontypool3/5 Peebles and South Midlothian +2/5 Roxburgh and Selkirk Pontypool Pontypool Torfaen 3/5 Barry + 3/5 Pontypridd 3/5 Barry + 3/5 Pontypridd 3/5 Barry + 3/5 Pontypridd Vale of Glamorgan West 3/5 Peebles and South Midlothian +2/5 Roxburgh and Selkirk + 1/5 Monmouth + 3/5 Pontypridd Base: 1997-2004 1945-1949 1950-1954 1955-1973 1974-1982 1983-1996 Lauderdale Scotland Tweeddale, Ettrick and Wales Torfaen Pontypool Wales Vale of ClwydWales Vale of GlamorganScotland WesternWales IslesWales 1/5 Denbigh Wrexham + 2/5 Flint 3/5 Llandaff and Ynys Barry Mon 1/5 Denbigh + 2/5 Flint Western Isles Wrexham Anglesey Western Isles Wrexham Anglesey Western Isles Wrexham Anglesey Western Isles Wrexham Western Isles Ynys Mon Wrexham Ynys Mon D Variables Description 76

D Variables Description

Nationalist vote share Share of votes received by SNP/Plaid Cymru in UK Parliament elections in a single constituency (in percent).

Discoveries (giant) Number of giant oil discoveries in a particular year t. An oil ﬁeld classiﬁed as ‘giant’ contains ultimate recoverable reserves of 500 million barrels or more before the extraction starts.

Discoveries (all) Number of oil discoveries in a particular year t. All oil ﬁelds with 50 million barrels or more are captured.

Amount of new reserves Reserves of discovered oil ﬁelds in year t in 1000 million barrels of oil (MMstb.).

Scotland Binary variable indicating Scottish constituencies (1 if the constituency is Scottish, 0 otherwise).

Oil pricea Real price of Brent crude oil (year average). The unit is constant 2001-US-$.

GDP per capita Relative regional per capita gross domestic product for Scotland and Wales (in percent of UK average).

Unemployment rate Regional rate of registered unemployed (Claimant count) for Scotland and Wales (in percent).

Near border (50)b Binary indicator for constituencies that are less than 50 km of the En- glish border (1 if the constituency is within this distance, 0 otherwise).

Near border (75)b Binary indicator for constituencies that are less than 75 km of the En- glish border (1 if the constituency is within this distance, 0 otherwise).

Near border (100)b Binary indicator for constituencies that are less than 100 km of the En- glish border (1 if the constituency is within this distance, 0 otherwise).

Coastal accessb Binary indicator for constituencies with coastal access (1 if the constituency has sea access, 0 otherwise).

Distance to Aberdeenb Distance from a constituency to Aberdeen (in km). D Variables Description 77

Avg. soil suitabilityc Average soil suitability for production of potatoes, barley, and wheat. (medium input intensity and irrigation).

Ruggedness indexd Index of variance of elevation in each constituency.

a To calculate the real oil price, we used US inﬂation data from the Bureau of Labor Statistics (see the data at http://www.usinfl ationcalculator.com/, last accessed on 19th August 2015). The Brent price prior to 1957 is approximately projected using data for the sort WTI. b Variables are calculated using ArcGIS. Data are taken from http://www.gadm.org, Boundary Commission for Scotland (http://www.bcomm-scotland.independent.gov.uk/maps/datafiles/), David Boothroyd (http://www.election.dem on.co.uk/), UK Data Service (https://census.edina.ac.uk/easy_download.html); all last accessed on 19th August 2015. c Data are taken from the Global Elevation Data Set (http://diegopuga.org/data/rugged/). d Data are taken from http://www.fao.org/nr/gaez/en/. E Estimated Scottish Shares 78

E Estimated Scottish Shares

Scottish Shares of Total UK Oil Production Source: (Kemp & Stephen, 2000)

The estimations by (Kemp & Stephen, 2000) are based on the assumption of the equidistance line as the maritime border; that is, “a dividing line on which all points are the same distance from the Scottish and rest of the UK coastline” (Brocklehurst, 2013). Another possible border would be the 55’50’ latitude, established for juristical reasons in 1968 (Brocklehurst, 2013; Lee, 1976). Alexander G. Kemp remarks that “from the economic point of view, it does not make much diﬀerence because there are just a handful of ﬁelds, and not very important ones now, between the median line and the line north of Berwick” (Brocklehurst, 2013). In addition to that: “[t]hese considerations aside, there is no doubt that most of the oil lies in the northern North Sea. However, two-thirds of known reserves are 100 miles east of the Shetland Islands and can morally be claimed by their inhabitants” (Lee, 1976, 310). F Regressions 79

F Regressions

Table 1: Regression Results – Alternative Clustering (Table 2) Nationalist Nationalist Nationalist Nationalist Nationalist Dependent variable vote share vote share vote share vote share vote share Scotland × Discoveries (giant) - 3.261 2.862 1.923 1.926 [0.304] [0.290] [0.332] [0.335] Scotland - 2.263 −3.500 - - [1.129] [0.508] Discoveries (giant) 4.494 - - - - [0.253] p-value: Scotland × Disc. (giant) - 0.000 0.000 0.000 0.000 Biannual fixed effects no yes yes yes yes Constituency-fixed effects yes no no yes yes Linear time trend Scotland no no yes yes no Constituency-specific time trends no no no no yes Adj. R-squared 0.58 0.50 0.52 0.74 0.83 Number of observations 1216 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. It corresponds to Table 2 in the paper but standard errors are clustered on the constituency level using the ivreg2 command in Stata. Discoveries (giant) denotes the number of giant oil fields discovered in t=0 and t=-1.

Table 2: Regression Results – Alternative Clustering (Table 2) Nationalist Nationalist Nationalist Nationalist Nationalist Dependent variable vote share vote share vote share vote share vote share Scotland × Discoveries (giant) - 3.261 2.862 1.923 1.926 [0.823] [0.781] [0.870] [0.868] Scotland - 2.263 −3.500 - - [1.429] [1.434] Discoveries (giant) 4.494 - - - - [1.124] p-value: Scotland × Disc. (giant) - 0.000 0.000 0.027 0.027 Biannual fixed effects no yes yes yes yes Constituency-fixed effects yes no no yes yes Linear time trend Scotland no no yes yes no Constituency-specific time trends no no no no yes Adj. R-squared 0.58 0.50 0.52 0.74 0.83 Number of observations 1216 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. It corresponds to Table 2 in the paper but standard errors are clustered on the biannual level using the ivreg2 command in Stata. Discoveries (giant) denotes the number of giant oil fields discovered in t=0 and t=-1. F Regressions 80

Table 3: Regression Results – Alternative Clustering (Table 2) Nationalist Nationalist Nationalist Nationalist Nationalist Dependent variable vote share vote share vote share vote share vote share Scotland × Discoveries (giant) - 3.261 2.862 1.923 1.926 [0.605] [0.574] [0.640] [0.638] Scotland - 2.263 −3.500 - - [1.051] [1.055] Discoveries (giant) 4.494 - - - - [1.124] p-value: Scotland × Disc. (giant) - 0.000 0.000 0.003 0.003 Biannual fixed effects no yes yes yes yes Constituency-fixed effects yes no no yes yes Linear time trend Scotland no no yes yes no Constituency-specific time trends no no no no yes Adj. R-squared 0.58 0.50 0.52 0.74 0.83 Number of observations 1216 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. It corresponds to Table 2 in the paper but standard errors are clustered on the country × time level using the ivreg2 command in Stata. Discoveries (giant) denotes the number of giant oil fields discovered in t=0 and t=-1.

Table 4: Regression Results – Alternative Clustering (Table 2) Nationalist Nationalist Nationalist Nationalist Nationalist Dependent variable vote share vote share vote share vote share vote share Scotland × Discoveries (giant) - 3.261 2.862 1.923 1.926 [0.605] [0.574] [0.140] [0.132] Scotland - 2.263 −3.500 - - [1.051] [1.055] Discoveries (giant) 4.494 - - - - [1.422] p-value: Scotland × Disc. (giant) - 0.000 0.000 0.000 0.000 Biannual fixed effects no yes yes yes yes Constituency-fixed effects yes no no yes yes Linear time trend Scotland no no yes yes no Constituency-specific time trends no no no no yes Adj. R-squared 0.58 0.50 0.52 0.74 0.83 Number of observations 1216 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. It corresponds to Table 2 in the paper but standard errors are clustered on the country × time-category level using the ivreg2 command in Stata with 5 successive time-categories. Discoveries (giant) denotes the number of giant oil fields discovered in t=0 and t=-1. F Regressions 81

Table 5: Regression Results – Alternative Clustering (Table 2) Nationalist Nationalist Nationalist Nationalist Nationalist Dependent variable vote share vote share vote share vote share vote share Scotland × Discoveries (giant) - 3.174 2.705 1.864 1.865 [1.189] [0.862] [0.826] [0.817] Scotland - 2.317 −3.461 11.479 −3.173 [1.436] [1.794] [4.565] [2.805] Discoveries (giant) 4.495 −1.450 2.157 10.83 −15.673 [1.096] [3.149] [3.721] [4.001] [9.341] p-value: Scotland × Disc. (giant) - 0.008 0.002 0.025 0.022 Biannual fixed effects no yes yes yes yes Constituency-fixed effects yes no no yes yes Linear time trend Scotland no no yes yes no Constituency-specific time trends no no no no yes Number of observations 1152 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. It corresponds to Table 2 in the paper but the estimation uses panel-corrected standard errors with panel-specific auto-correlation. Discoveries (giant) denotes the number of giant oil fields discovered in t=0 and t=-1.

Table 6: Regression Results – Alternative Clustering (Table 2) Nationalist Nationalist Nationalist Nationalist Dependent variable vote share vote share vote share vote share Scotland × Discoveries (giant) 1.923 1.926 1.923 1.926 [0.640] [0.638] [0.140] [0.132] Bootstrap p-value (2-point): Scotland × Disc. (giant) 0.100 0.086 0.068 0.066 Biannual fixed effects yes yes yes yes Constituency-fixed effects yes yes yes yes Linear time trend Scotland yes no yes no Constituency-specific time trends no yes no yes Adj. R-squared 0.74 0.83 0.74 0.83 Number of observations 1883 1883 1883 1883 The table displays regression coefficients with standard errors in brackets. Estimations correspond to the two last columns in Table 2 in the paper. Standard errors are clustered on the country × time level (in the left two columns) and the country × time-category level (in the right two columns) using the ivreg2 command in Stata. Bootstrap p-value refers to p-values estimated with two wild-cluster bootstrap procedures (using a 2-point distribution) with 10,000 repetitions. Discoveries (giant) denotes the number of giant oil fields discovered in t=0 and t=-1. REFERENCES 82

References

Ashbrook, John, & Ashbrook, John E. 2008. Buying and Selling the Istrian Goat: Istrian Regionalism, Croatian Nationalism, and EU Enlargement. Peter Lang. 60

Brocklehurst, Steven. 2013. Who Has a Right to Claim North Sea Oil? BBC Scotland News, http: //www.bbc.co.uk/news/uk-scotland-scotland-politics-20042070. 78

Casey, B. J., Romani, R. S., & Schmitt, R. H. 1993. Appraisal Geology of the Saltire Field, Witch Ground Graben, North Sea. Geological Society, London, Petroleum Geology Conference series. 66

Coward, R. N., Clark, N. M., & Pinnock, S. J. 1991. The Tartan Field, Block 15/16, UK North Sea. Geological Society, London, Memoirs, 14(1), 377–384. 66

Deschouwer, Kris. 2013. Party Strategies, Voter Demands and Territorial Reform in Belgium. West European Politics, 36(2), 338–358. 44, 45

Duerr, Glen M. E. 2016. Second-Tier Secessionism in Europe: How Has Support for Independence in Wales, Wallonia, and Galicia Changed with Increasing Claims in Scotland, Flanders, and Catalonia? History and Government Faculty Presentations No. 162. 47, 50

Eneyok, G., Bussink, P., & A., Maan. 2003. The Curlew Field, Block 29/7, UK North Sea. In: United Kingdom Oil and Gas Fields: Commemorative Millennium Volume. Gluyas J.G. and Hichens H.M. 66

EnQuest. Heather/Broom. http://www.enquest.com/our-business/operated-assets/heather-b room.aspx. 66

EnQuest. 2013 (October). Enquest Acquires 50% of the Greater Kittiwake Area. http://www.enquest.co m/media-centre/press-releases/2013/kittiwake.aspx. 66

E.ON. 2013 (April). Successful Start of Production from Huntington. https://www.eon.com/en/media/n ews/press-releases/2013/4/15/successful-start-of-production-from-huntington.html. 66

Favero, Carlo A., Hashem Pesaran, M., & Sharma, Sunil. 1994. A Duration Model of Irreversible Oil Investment: Theory and Empirical Evidence. Journal of Applied Econometrics, 9(S1), S95 – S112. 66

Fee, Derek A, & O’Dea, John. 2005. Technology for Developing Marginal Oﬀshore Oilﬁelds. CRC Press. 66

Giordano, Benito. 2000. Italian Regionalism or Padanian Nationalism-the Political Project of the Lega Nord in Italian Politics. Political Geography, 19(4), 445–471. 63 REFERENCES 83

Glennie, K. W., & Armstrong, L. A. 1991. The Kittiwake Field, Block 21/18, UK North Sea. Geological Society, London, Memoirs, 14(1), 339–345. 66

Guscott, Simon, Russell, Ken, Thickpenny, Andrew, & Poddubiuk, Robert. 2003. The Scott Field, Blocks 15/21a, 15/22, UK North Sea. Geological Society, London, Memoirs, 20(1), 467–482. 66

Hermans, Theo. 2015. The Flemish Movement: a Documentary History 1780-1990. Bloomsbury Publishing. 44

Horowitz, Leah S. 2004. Toward a Viable Independence? The Koniambo Project and the Political Economy of Mining in New Caledonia. The Contemporary Paciﬁc, 16(2), 287–319. 56

Jayasekera, A. J., Goodyear, S. G., et al. 1999. The Development of Heavy Oil Fields in the UK Continental Shelf: Past, Present and Future. In: SPE Western Regional Meeting. Society of Petroleum Engineers. 66

Kavanagh, Michael. 2013. Total Resumes Elgin-Franklin Production. http://www.ft.com. 66

Kay, Simon. 2003. The Heather Field, Block 2/5, UK North Sea. In: United Kingdom Oil and Gas Fields: Commemorative Millennium Volume. Gluyas J.G. and Hichens H.M. 66

Kemp, Alexander G., & Stephen, Linda. 2000. Expenditures in and Revenues from the UKCS: Estimating the Hypothetical Scottish Shares. North Sea Study Occasional Paper No. 70. 78

Kunka, J. M., Williams, G., Cullen, B., Boyd-Gorst, J., Dyer, G. R., Garnham, J. A., Warnock, A., Wardell, J., Davis, A., & Lynes, P. 2003. The Nelson Field, Blocks 22/11, 22/6a, 22/7, 22/12a, UK North Sea. Geological Society, London, Memoirs, 20(1), 617–646. 66

Lee, Geoﬀrey W. 1976. North Sea Oil and Scottish Nationalism. Political Quarterly, 47(3), 307–317. 78

Muñoz, Jordi, & Tormos, Raül. 2012. Economic Expectations and Preferences for Secession: Causality or Rationalization? Working Paper. 49

Nexen/CNOOC. 2013 (October). Golden Eagle. http://www.nexencnoocltd.com/en/Operations/Co nventional/UKNorthSea/GoldenEagle.aspx. 66

Pye, M., & Brown, S. 2002. Hydrocarbons. Geological Society, London, The Geology of Scotland. 66

Reid, Alasdair, & Musyck, Bernard. 2000. Industrial Policy in Wallonia: A Rupture With the Past? European Planning Studies, 8(2), 183–200. 47

Ritchie, J. S. 2003. The Dunbar, Ellon and Grant Fields (Alwyn South Area), Blocks 3/8a, 3/9b, 3/13a, 3/14, 3/15, UK North Sea. In: United Kingdom Oil and Gas Fields: Commemorative Millennium Volume. Gluyas J. G. and Hichens H.M. 66 REFERENCES 84

Steblez, Walter G. 2005. The Mineral Industries of Central Europe. US Geological Survey Minerals Yearbook. 57

Talisman Energy. 2006a. Claymore. https://www.talisman-energy.com. 66

Talisman Energy. 2006b. Clyde Platform. http://www.talisman-energy.com. 66

Talisman Energy. 2007. Tartan Alpha. http://www.talisman-energy.com. 66

The Maersk Group. 2014 (May). Where Others gave up Maersk Oil Got 40 Million Bar- rels. http://www.maersk.com/en/hardware/innovation/case-stories/where-others-gave-u p-maersk-oil-got-40-million-barrels. 66

Toal, Gerard. 2013. Republika Srpska Will Have a Referendum: the Rhetorical Politics of Milorad Dodik. Nationalities Papers, 41(1), 166–204. 57

Tonkin, P. C., & Fraser, A. R. 1991. The Balmoral Field, Block 16/21, UK North Sea. In: United Kingdom Oil and Gas Fields: 25-Years Commemorative Volume. Gluyas J. G. and Hichens H.M. 66

United Kingdom Government. 2013. Appendix 2 - Historic UK Oil and Gas Reserves and Production Used In Reserves Charts v Time for Web Site. https://www.gov.uk/government. 66

Van Vessem, E.J., & Gan, T.L. 1991. The Ninian Field, Blocks 3/3 & 3/8, UK North Sea. Geological Society, London, Memoirs, 14(1), 175–182. 66

Walker, Raymond (Sun Oil Britain Ltd.). 1994. The Balmoral Field Subsea Development: A Case History. In: Oﬀshore Technology Conference, 2.-5. May, Houston, Texas. 66

Witte, Els. 1992. Belgian Federalism: Towards Complexity and Asymmetry. West European Politics, 15(4), 95–117. 47 G It’s Scotland’s Oil! 85

G It’s Scotland’s Oil!

Poster from the SNP’s “It’s Scotland’s Oil” campaign in the 1970s